Started in January,1974(Monthly)
Supervised and Sponsored by Chongqing Southwest Information Co., Ltd.
ISSN 1002-137X
CN 50-1075/TP
CODEN JKIEBK
Editors
Current Issue
Volume 44 Issue 11, 01 December 2018
  
Research Status and Prospect of Image Dehazing
MIAO Qi-guang and LI Yu-nan
Computer Science. 2017, 44 (11): 1-8.  doi:10.11896/j.issn.1002-137X.2017.11.001
Abstract PDF(1532KB) ( 895 )   
References | Related Articles | Metrics
Images captured under haze are always degraded due to the suspending aerosols in air,which can also affect their applications in the outdoor computer vision system.Then researchers proposed a lot of algorithms for image restoration.Current image dehazing algorithms were analyzed,including two categories of image enhancement based and physical model based.Then the advantages and disadvantages of several highly applied algorithms were discussed,and some quantity evaluation methods were introduced.Finally,some possible study ways and the trend of development were given.
Automatic Generation and Evolution of Policies for Self-adaptive Software
LIN Hua-shan, LIU Yang and JIAO Wen-pin
Computer Science. 2017, 44 (11): 9-14.  doi:10.11896/j.issn.1002-137X.2017.11.002
Abstract PDF(885KB) ( 497 )   
References | Related Articles | Metrics
With the increment of complexities of the software features and uncertainties in the software runtime environment,software requires the ability to perceive and react to the changes from the environment in time.The self-adaptive system is a software framework with the ability to choose actions according to environment and knowledge.The set of policies plays a key role in the knowledge of self-adaptive system,but there’s little research which focus on producing,management and evolution of policies.Based on the decision tree algorithm,a method of producing,management and evolution of policies was proposed,which makes the self-adaptive software perform well in using knowledge and avoiding the detrimental effect from drawing up the policies artificially.At the same time,an experiment,which is based on the RubiS website,was conducted to prove the feasibility of this method.
Study on Technical Debt Management of Integrated Development Environment
LIU Ya-jun, LI Bing, LI Zeng-yang, LIANG Peng and WU Min-quan
Computer Science. 2017, 44 (11): 15-21.  doi:10.11896/j.issn.1002-137X.2017.11.003
Abstract PDF(1291KB) ( 575 )   
References | Related Articles | Metrics
Technical debt (TD),a metaphor to financial debt,refers to technical compromises which are made to gain short-term benefits at the cost of long-term software quality.However,in the long run TD will negatively affect software quality as well as cost and productivity of software development.Thus,TD should be effectively managed.Existing dedicated TD management tools are rather limited and have various limitations.Hence,it is difficult to effectively mana-ge TD using such tools.Mainstream integrated development environments (IDEs) are powerful and widely used,can serve to manage TD.In this study,we investigated Microsoft Visual Studio 2015 Enterprise Edition,a representative mainstream IDE,to explore its ability for managing four types of TD directly related to source code using a C# project.Then we compared Visual Studio with four dedicated TD management tools with respect to their capabilities in mana-ging TD and their support for TD management of development teams in their daily practices.The results show that VisualStudio provides better support for managing TD in terms of diversity of TD types and the range of TD management activities.In addition,with Visual Studio,specific TD instances can be managed to different extents in a variety of ways.
JavaScript Counterfactual Execution Method Based on Dynamic Instrumentation
GONG Wei-gang, YOU Wei, LI Zan, SHI Wen-chang and LIANG Bin
Computer Science. 2017, 44 (11): 22-26.  doi:10.11896/j.issn.1002-137X.2017.11.004
Abstract PDF(792KB) ( 659 )   
References | Related Articles | Metrics
The static analysis technique has been widely employed in the security analysis of JavaScript program.But the JavaScript program can leverage several functions such as eval to generate code at runtime,which is hard to obtain danamic generation code simply by static analysis.One feasible approach is to collect the code by running the target program dynamically and then make a static analysis on it.However,this approach can only explore a finite number of execution paths and will miss the dynamically generated code in other paths.This paper presented a counterfactual execution method based on dyna-mic instrumentation.In the method,the counterfactual execution structures are instrumented on-the-fly during the parse phase of JavaScript engine,to explore both the branch that would ordinarily be executed and the other branch that would not normally be run.In this way,even if the functions like eval are called nestedly,the dynamically generated code can also be instrumented.Besides,in order to undo the effect of any assignment in counterfactual execution structures,an on-demand undo method was implemented to avoid the redundant operations.The evaluation results show that the method implemented in this paper can effectively expand the coverage of execution paths in dynamic analysis.
Concurrency Bugs Verification in Android Applications Based on Test Case Generation
SHE Yi, TANG Hong-yin, WU Guo-quan, CHEN Wei, WEI Jun and HUANG Tao
Computer Science. 2017, 44 (11): 27-32.  doi:10.11896/j.issn.1002-137X.2017.11.005
Abstract PDF(908KB) ( 567 )   
References | Related Articles | Metrics
Android application is an event-based concurrent program.Data race is common in Android apps due to unforeseen threads interleaving coupled with non-deterministic reordering of asynchronous events.Existing race detection tools for Android applications will report false positive,and cannot replay the concurrent bugs caused by race.To address this issue,this paper proposed a new technique to expose race based on the results obtained from existing race detection tools.Our technique firstly builds state transition graph (STG) for targeted application,and generates a test case that has potential data races based on the STG and the results reported by existing race detection tools.Then it resche-dules test cases execution by controlling event dispatching and thread interleaving to determine whether such potential races really lead to thrown exceptions.Our experiments show that this technique is effective,and it can confirm and replay concurrency bug caused by real data races,while at the same time eliminates false warnings for Android apps.
Cloud Operating System for Cloudware
WANG Wei, HU Chang-wu, GUO Dong, ZHANG Jing-xuan, CHANG Jin-da and ZHANG Li-qing
Computer Science. 2017, 44 (11): 33-40.  doi:10.11896/j.issn.1002-137X.2017.11.006
Abstract PDF(1310KB) ( 702 )   
References | Related Articles | Metrics
The rise of cloud computing and the Internet not only bring change on the data center,but also lead to transformation in software development,deployment,operation and maintenance.With the continuous improvement of the current cloud computing and the Internet environment,how to make better use of cloud computing platforms,and how to serve the end users in the field of computer software are big challenges.In recent years,with the further development of concepts like micro-services and containers,software will further step forward to the Cloudware.This paper defined the concept of Cloudware paradigm,and discussed how to deploy Cloudware in cloud environment.Then,we proposed a new cloud operation system called GalaxyOS which can directly deploy software on the cloud without any modification,while achieving a new model by the browser services.By using micro-service architecture,we can achieve such characteristics as good performance,scalable deployment,faults tolerance and flexible configuration.Finally,we evaluated Ga-laxyOS by proposing a user experience oriented framework and carrying out an interactive delay experiment that directly focuses on users’ experience,and verified the effectiveness of our platform.
Empirical Study of Reopened Security Bugs on Mozilla
ZHANG Kai, SUN Xiao-bing, PENG Xin and ZHAO Wen-yun
Computer Science. 2017, 44 (11): 41-49.  doi:10.11896/j.issn.1002-137X.2017.11.007
Abstract PDF(1378KB) ( 527 )   
References | Related Articles | Metrics
Compared to other types of bugs,security bug reopens more often,moreover,they need more development resources to fix it,which adds an extra cost to fix them.Hence,the empirical study of reopened security bugs is important.Our study collected the reopened security bugs from the Mozilla project,and analyzed them from the times of their reopening and commits,files which were modified to fix them,lines of added and deleted code,and comparison of the original fixing and reopened fixing.The empirical results show that security bug reopening often happen and it relates to the complexity of recognizing the reason that a security bug happens and fixing bugs.In addition,the locality of the files and code in the original security bug fixing is one of the causes to influence its re-fixing for bug reopens,and using more complex and effective fixing process can help reduce the security bug reopens.Finally,we summarized several causes for security bug reopens to help developers more easily identify the reopens of different types of security bugs.
Sandstone Microscopic Image Analysis Method and Tool Implementation
HAO Hui-zhen, JIANG Feng, LI Na and GU Qing
Computer Science. 2017, 44 (11): 50-55.  doi:10.11896/j.issn.1002-137X.2017.11.008
Abstract PDF(1039KB) ( 798 )   
References | Related Articles | Metrics
Image analysis is an important method for studying sandstone.The research to develop methods which are suitable for sandstone microscopic image analysis and its implementation are valuable for both studying sandstone petrology and oil-gas exploration.This work developed a software system for sandstone microscopic image analysis.Firstly,superpixel segmentation method SLIC is adapted to segment microscopic images of sandstone which forms superpi-xels with only one mineral ingredient.Secondly,as the training data,the color and local features are extracted from micro-mineral images,and are used to train classifier to classify superpixels.Lastly,those adjective superpixels are merged to a whole mineral grain which is labeled with its category as a result.Based on this method,a set of tools were designed to perform mineral composition and texture analysis on the sandstone microscopic images.The analysis on microscopic images of sandstones from Tibet verified this method to be practical and useful.However,this developed software tool need to be further improved and optimized.
Multi-layer Model for SaaS Multi-tenancy
LI Xiang, LI Tong, XIE Zhong-wen, HE Yun, CHENG Lei and HAN Xu
Computer Science. 2017, 44 (11): 56-63.  doi:10.11896/j.issn.1002-137X.2017.11.009
Abstract PDF(1134KB) ( 725 )   
References | Related Articles | Metrics
SaaS(Software as a Service) appears with cloud computing.It is much different from traditional software.Based on the characters of SaaS,a hierarchical meta-model of SaaS was proposed,which supports the SaaS maturity model and describes every layer by formalization method.Inspired by object-oriented Petri nets (OOPN) and colored Petri nets(CPN),SOP(Service-Oriented Petri Nets) and CSOP(Colored Service-Oriented Petri Nets) were proposed.On the one hand,encapsulated place is used to represent service,which reflects that services are invisible to outside and inside influences running situation of system.On the other hand,different color sets represent the requests of different tenements,which emphasizes multi-tenancy character of SaaS.The work of this paper not only provides a function for modeling complex SaaS software,but also folds the transition of petri net and encapsulates state space of system.A CRM(Customer Relationship Management) system of SaaS was illustrated to verify the feasibility in the end.
Backporting of Linux Device Drivers Using Semantic Patch
ZHU Li-hua, WEN Yan-jun and DONG Wei
Computer Science. 2017, 44 (11): 64-68.  doi:10.11896/j.issn.1002-137X.2017.11.010
Abstract PDF(824KB) ( 695 )   
References | Related Articles | Metrics
The paper studied the backporting of network device drivers in Linux kernels using semantic patch technology.After analyzing the code of device drivers in many different kernel versions,this paper proposed an approach to backport device drivers using semantic patch technology with compatibility library.Semantic patches are developed sepa-rately according to symbols and functions, which reduces the redundancy of code and improves the efficiency of backporting.The experiment shows that the lines of code of semantic patches are much less than normal patches.Some experiences about building compatibility library were also proposed in the paper.The results show that this approach is also valuable for backporting other device drivers.
Method of Java Code Repository Construction Based on UML Relationship
JIANG Ren-he, ZHENG Xiao-mei, ZHU Xiao-qian, PAN Min-xue and ZHANG Tian
Computer Science. 2017, 44 (11): 69-79.  doi:10.11896/j.issn.1002-137X.2017.11.011
Abstract PDF(1575KB) ( 730 )   
References | Related Articles | Metrics
Relationship information is the most important information representing code structure and semantics,such as inheritance,aggregation,composition,dependency,call and creation.This paper provided a method constructing source repositories based on relationships defined in UML2 for better comprehension and reuse of the open source code.The graph database is used as the implementation plantform of the approach,and the abstract syntax tree is adopted as the base of graph schema.In addition,the schema is designed specifically according to the Java language so that semantics can be well presented.The key point of the approach is that the relationship definitions are prescribed strictly according to the UML2.4 specifications,which is an ISO version,so that the differentiation between different language communities about the comprehension of relationship can be eliminated.Each category of the relationships was studied and the corresponding recognition algorithms were designed.During the construction of the repository,the relationships are added as different kind of edges.The evaluation experiments of 9 open source projects were conducted to illustrate the expansion of the code graph and the consumed space.At last,two simple case studies of querying on the repository were given.
Dynamic Evolution Consistency Verification Method for Component System Modeling
ZHENG Ming, LI Tong, LIN Ying, ZHOU Xiao-xuan, LI Xiang and MING Li
Computer Science. 2017, 44 (11): 80-86.  doi:10.11896/j.issn.1002-137X.2017.11.012
Abstract PDF(1256KB) ( 541 )   
References | Related Articles | Metrics
Component-based technology has become a main stream approach for software development,however,how to maintain a high level of consistency of the component system after dynamic evolution is still lacking a generally accepted standard.So a verification method was proposed to verify whether a component system is consistent after dynamic evolution.Firstly,component model was proposed based on process algebraic,and a complex component system model could be obtained by combination of the components model.Secondly,according to the component system model and its state changes,an algorithm for extracting the external behavior of the component system was proposed,and a verification standard of consistency of the component system was presented based on weak bisimulation theory.Finally,the behavior of the component system was obtained,and then these behaviors were converted into a format which is conve-nient for Mobility Workbench to identify and verify.The case study shows that the proposed approach is feasible and effective.
Improved Method of Context-sensitive Control Flow Integrity Protection
SHEN Qin-tao, ZHANG Li, LUO Lei, MA Jun, YU Jie and WU Qing-bo
Computer Science. 2017, 44 (11): 87-90.  doi:10.11896/j.issn.1002-137X.2017.11.013
Abstract PDF(571KB) ( 559 )   
References | Related Articles | Metrics
Facing the threat of control flow hijacking,the industry uses control flow integrity protection technology.It is difficult to achieve the goal for those traditional control-flow integrity protection mechanisms which depend on dynamic binary rewriting technology,and it’s not easy for analysis and implementation.It may also bring out the problem of binary compatibility.The recently proposed context-sensitive control-flow integrity,PathArmor,only verifies the control flow when tasks are calling system functions.To achieve enhanced protection,an improved method was proposed in this paper.The improved method means to trigger more page fault intentionally by modifying the pages’ protection flag of the target task,with the kernel’s mechanism of page fault.Then it hooks the origin system IDT (Interrupt Description Table) and creates new do_page_fault function to handle the generated page fault.With doing some experiments on typi-cal application like nginx,bzip2,SQLite and so on,the result shows that the counts for tasks to be verified increase significantly,and it can get better protection with the improved method.
Deriving Invariants from Database Transactions
ZENG Hong-chi, PENG Xin and ZHAO Wen-yun
Computer Science. 2017, 44 (11): 91-97.  doi:10.11896/j.issn.1002-137X.2017.11.014
Abstract PDF(992KB) ( 532 )   
References | Related Articles | Metrics
As the basic unit of data processing and concurrency control,database transaction is widely used in business logic of software applications.It is a significant method in software maintenance for monitoring internal state to derive invariants and establish corresponding data contracts based on data collected from database transactions at runtime.Currently,in the field of invariant derivation,most researches and tools are just based on deriving invariants from source code, while there is lack of researches on deriving invariants from data.To solve this problem,this paper firstly introduced an algorithm for deriving invariants in the form of algebraic equations.Moreover,a prototyping tool was designed and implemented to collect data from database transactions at runtime and derive invariants among these data.With experiments,the prototyping tool is proved to be effective both on accuracy and performance,which fills the gap of inva-riants derivation based on data.
Document Based Matching Method for Mobile UI Components
XU Tong-tong, LIU Qu-tao, ZHENG Xiao-mei, PAN Min-xue and ZHANG Tian
Computer Science. 2017, 44 (11): 98-103.  doi:10.11896/j.issn.1002-137X.2017.11.015
Abstract PDF(917KB) ( 554 )   
References | Related Articles | Metrics
Development for multi-platform is an important requirement of developing mobile applications.At the same time,the developments should be flexible for short cycle and rapid evolution.These are challenges for mobile developments.Fortunately,most mobile platforms are designed as MVC-based and event/UI-driven.Hence the UI components are similar between different platforms,which is very helpful when developing Apps from one platform to another.This paper provided a document based matching method for uncovering the similarities of UI components between different platforms.The iOS and Android are selected,and the documentations for their UI components are extracted from the official websites.Then the NLP techniques are used to build the vector space model so as to compute the similarities of UI components between two platforms.To increase the accuracy,the sets of synonymous words were presented according to the UI features of components.The experiments were performed on a set of typical iOS and Android UI components.The results illustrate that the accuracy of the method is acceptable for the most UI components especially for those of one-one-matching.
Study of Component Assembling Technologies under Object-oriented Trusted Component Model
ZHAO Na, WANG Jian, LI Tong, YU Yong, LI Peng and XIE Zhong-wen
Computer Science. 2017, 44 (11): 104-108.  doi:10.11896/j.issn.1002-137X.2017.11.016
Abstract PDF(670KB) ( 424 )   
References | Related Articles | Metrics
The component-based trusted software construction and its key theories and technologies are some of the research focuses in the domain of the trusted software.Our previous work proposed a trusted component net model based on extended object-oriented Petri net so as to provide a set of description tools for component evolution.This paper studied the component assembling techniques under our trusted component net model.Components are assembled together in a meaningful and organic way by means of connectors.The definitions of four types of connector-based components assembling approaches,named parallel assembling,selective assembling,sequential assembling,repetitive assembling,were given,and subsequently their properties and proofs were detailed.
Mutation Test Based Test Case Minimization for Concurrent Program
ZHENG Wei, FENG Chen, WU Xiao-xue, HUANG Yue-ming and FANG Jing-yun
Computer Science. 2017, 44 (11): 109-113.  doi:10.11896/j.issn.1002-137X.2017.11.017
Abstract PDF(667KB) ( 436 )   
References | Related Articles | Metrics
In the test of concurrency program,test input and the timing of threads interaction are the two key factors that affect the concurrency bug detection.For narrowing input space of concurrency bug detection,this paper proposed an algorithm of test case minimization based on mutation test.It first studies on different concurrency programs and selects 9 mutation operators which can contribute to the generation of mutants to the testing programs.Then it chooses JPF as the detecting tool during the execution of programs,through which the test cases are sorted according to mutation score(MS) and average cost(AC).Finally it picks those test cases which can find out different bugs in the prioritized test cases.Experimental results show that the method can effectively reduce the size of test cases,shorten running time and improve the effectiveness of concurrency bug detection.
Design and Implementation of Automatic Code Generator for TLS System
WANG Jia-long, LIU Yan-hong and SHEN Li
Computer Science. 2017, 44 (11): 114-119.  doi:10.11896/j.issn.1002-137X.2017.11.018
Abstract PDF(824KB) ( 514 )   
References | Related Articles | Metrics
Although thread level speculation (TLS) mechanism can simplify the interface of multi-thread programming model and achieve high performance speedup,the development of its parallel program is still a tough task.Oriented to an effective software TLS model HEUSPEC,the design and implementation of automatic code generator C2H,which has three components.First,simple directives are designed for HEUSPEC,which indicates some key characteristics of parallel regions.Second,an algorithm is proposed to convert directives and parallel regions to speculative functions.At last,an algorithm is designed to generate HEUSPEC parallel codes.This method has been implemented on an open source compiler——Clang.Experimental results oriented to typical benchmarks from Rodinia and OmpSrc indicate that C2H can convert serial C codes with simple directives to HEUSPEC parallel codes,and the performance of automatic-generated parallel codes is very close to the codes developed by programmers manually.
Research,Design and Implementation of Android Multi-window System
LUO Hao, WANG Dan, XIAO Luo-yuan and CHEN Yu
Computer Science. 2017, 44 (11): 120-124.  doi:10.11896/j.issn.1002-137X.2017.11.019
Abstract PDF(867KB) ( 1049 )   
References | Related Articles | Metrics
Multi-window technology which is regarded as a way to improve the man-machine interface,has been popular in design of desktop system.As a phone system,Android has a large amount of user groups,and gradually expands to the large-screen devices.However,Current Android systems lack supports from multi-window,which greatly restricts users to use the Android system in the large-screen environment.Based on Android,focusing on the desktop platform,a multi-window system was designed and implemented for Android through the analysis for the functional requirements of the Android multi-window system.The availability and practical significance of Android multi-window system was assessed by displaying the effects and performance of the system.
Research on Quality Evaluation Approaches for Search Function of Online Shopping Platforms
CHEN Hao, TAO Chuan-qi and WEN Wan-zhi
Computer Science. 2017, 44 (11): 125-133.  doi:10.11896/j.issn.1002-137X.2017.11.020
Abstract PDF(1332KB) ( 1190 )   
References | Related Articles | Metrics
With the rapid development of big data,various big data applications are now in service in different fields.Typical big data applications,such as online shopping and retailing platform,face recognition system,intelligent decision system,self-help service system,medical treatment system make daily life more convenient.Search system is one of the most used big data applications.However,search system varies in different platforms,and there are few standards for it.The quality of search system is hard to assure and validate.Search engine for online shopping systems combines text search and classification-based retrieval comparing to common text search engines.It is harder to validate and evaluate quality of it.Through studying on search system of online shopping platform,some quality factors and relational implementation algorithm were provided in this paper to validate and evaluate shopping system search engines.Experiments were also carried out to assure the correctness of the quality index.
Mobile Application Security Policies and Testing Research on XACML
CAO Wan-tian and YU Peng-fei
Computer Science. 2017, 44 (11): 134-145.  doi:10.11896/j.issn.1002-137X.2017.11.021
Abstract PDF(1749KB) ( 828 )   
References | Related Articles | Metrics
With the development of mobile Internet technology,the mobile terminals that have the ability to compute are deployed in great quantities.They can complete various tasks with the support of a large number of mobile applications.More and more companies allow employees to bring their own devices into the work environment,and this can be called BYOD (Bring Your Own Device).But different people have different characters,and different resources have different access permissions.The leak of sensitive resources will lead to significant losses of the enterprise.If BYOD wants to be supported perfectly,it is important to ensure the security of data and system.The access control rules that are defined for access to sensitive resources from the corresponding mobile applications need to be clearly and to be implemented in the running process of mobile applications.XACML is an unified description language of access control policies.Until now,it is unable to support mobile applications and BYOD.In this paper,we proposed a study method of testing XACML policies based on that XACML can describe access control policies of mobile applications.We conducted a case studywith a project management app facing BYOD on the Android platform and showed the validity of our method.
Prediction on Closed-probability of Change Request Report for Issue Tracking System
XIONG Wen-jun, ZHANG Xuan, WANG Xu, LI Tong and YIN Chun-lin
Computer Science. 2017, 44 (11): 146-155.  doi:10.11896/j.issn.1002-137X.2017.11.022
Abstract PDF(1409KB) ( 473 )   
References | Related Articles | Metrics
There are lots of change request reports without being closed for a long time in the Issue tracking system,which increases the likelihood of developers to click and read the reports again and again.It seriously affects the imple-mentation of management task of software requirements and the feedback experiences of users.The accurate and instant prediction of closed-probability or importance of these reports can improve the quality of the task of software maintenance.Several metrics were defined to measure the feature of change request,and Logistic regression prediction model was built by using these best predictive metrics on datasets for training.Then experiments which applied proposed method were performed on datasets of testing which contain 20 SourceForge projects,and achieved a result that average recall of 95% and average FPR (False Positive Rate) of 14%.Analysis of experimental result shows that the proposed method can achieve a good prediction performance on datasets of testing,and closed-percentage or size of change requests report doesn’t affect the performance of the model,and some features of change request report can be used to predict its closed-probability in the next version.
Random Search Rule Based Performance Evolutionary Optimization Method at Software Architecture Level
NI You-cong, LI Song, YE Peng and DU Xin
Computer Science. 2017, 44 (11): 156-163.  doi:10.11896/j.issn.1002-137X.2017.11.023
Abstract PDF(1261KB) ( 502 )   
References | Related Articles | Metrics
The existing rule-based performance optimization approaches at software architecture (SA) level don’t fully concern the improvement range of the rule and the uncertainty of the count and the order of each rule usage in the optimization process.As a result,the search space for performance improvement is limited and the better solutions are hard to find.Aiming at this problem,firstly,a group of the random search rules (RSRs) were designed based on performance improvement tactics so that each RSR can explore the larger performance improvement space.Then the combination of the order and the count of each rule usage was involved in the definition of the random search rule based performance optimization model named RRPOM and the evolutionary algorithm was designed to solve RRPOM.Further,the random search rule based method for performance optimization at software architecture level (RRMO4PO) was proposed.Finally,a WebApp application was taken as a case in the experiments for comparing RRMO4PO with the existing methods.The experimental results show that RRMO4PO can obtain the solutions with better interpretability by using fewer rules and fewer times to modify the SA elements.In addition,the results also prove that both system response time and cost for performance improvement can be decreased more efficiently in our approach.In the best results in our experiments,the average number of rules with improvement effect and the times to modify the SA elements are reduced by 33.3% and 52.9% respectively,and the system response time and the cost for performance improvement are decreased by 30.5% and 73.6% respectively.
Design and Implementation of Recommender System Based on Random Forest Algorithm
SHEN Jing-lei, YU Hui-qun, FAN Gui-sheng and GUO Jian-mei
Computer Science. 2017, 44 (11): 164-167.  doi:10.11896/j.issn.1002-137X.2017.11.024
Abstract PDF(712KB) ( 1288 )   
References | Related Articles | Metrics
As the recommender system is gaining momentum nowadays,how to quickly and accurately predict users’ behavior is becoming increasingly important.We implemented a recommendation system based on the random forest algorithm through analyzing the praise and stamp data collected from Internet forum.The system transforms the practical problem into a classification model,and it realizes data processing,feature extraction and parameter tuning.Moreover,the system determines whether there is a further interaction behavior after a post has been viewed.Finally,we conducted experiments to evaluate the system in terms of F1-measure.Our experimental results demonstrate the effectiveness and efficiency of our system.
Study on Invulnerability of Router-level Internet Based on MapReduce
ZHU Kai-long, LU Yu-liang and YANG Bin
Computer Science. 2017, 44 (11): 168-174.  doi:10.11896/j.issn.1002-137X.2017.11.025
Abstract PDF(1003KB) ( 447 )   
References | Related Articles | Metrics
We used MapReduce to study the invulnerability of router-level Internet.In order to evaluate network invulnerability,two new measures of invulnerability were proposed:network connectivity ratio and efficiency ratio.We designed the analysis algorithm of the Internet invulnerability based on MapReduce (AIIMR) using the two new mea-sures.The algorithm simulates the different attack on the Internet.The experimental results show the problem of traditional measure in measuring the router network and prove the validity of the proposed measure.The experimental results also show that the attacked router-level Internet is robust at random attack and fragile at targeted attack.Finally,the experiment proves that the proposed MapReduce algorithm effectively improves the efficiency,and has good sca-lability.
MBFL with Statement-oriented Mutant Reduction Strategy
WANG Lin-xin, WANG Wei-wei, ZHAO Rui-lian and LI Zheng
Computer Science. 2017, 44 (11): 175-180.  doi:10.11896/j.issn.1002-137X.2017.11.026
Abstract PDF(911KB) ( 563 )   
References | Related Articles | Metrics
How to efficiently and accurately locate faults in programs during the process of software debugging is taken up as a matter of common concern for software developers.MBFL is a fault localization technique based on mutation analysis,which precisely identifies the root cause of fault but incurs a high execution cost,since it needs to execute the test suite on a large amount of mutants.For decreasing the execution cost of MBFL,this paper presented a statement-oriented mutant reduction strategy,which selects a certain proportion set of mutants generated by statements covered by failed tests,according to the previous execution information of test suite.Empirical studies were conducted on 112 faulty versions from 7 program packages.The results indicate that this strategy can reduce 73.51%~79.98% mutation execution cost under the case of keeping high fault location precision.
Analytical Method for APP Software’s User Comment Patterns
RAN Meng and JIANG Ying
Computer Science. 2017, 44 (11): 181-186.  doi:10.11896/j.issn.1002-137X.2017.11.027
Abstract PDF(851KB) ( 693 )   
References | Related Articles | Metrics
Faced with the massive user’s comment of APP software,there are different results of comment focus,expression style and emotional tendency for different users.These differences bring difficulties to the user behavior analysis and the quality evaluation of the APP software.This paper presented a method to analyze APP software’s user comment patterns.At first,the relationships between user’s comment information and APP software information are analyzed.The user’s comment information is classified based on the comment feature of APP software.Then the combination of part-of-speech for each type of user’s comment information is analyzed.After calculating the emotional tendency degree of user’s comment information,the user’s comment patterns are analyzed.Finally,the experimental results show that the method is effective.
Research on Behavior-driven Development Based on Problem Frames
GAO Ning and LI Zhi
Computer Science. 2017, 44 (11): 187-190.  doi:10.11896/j.issn.1002-137X.2017.11.028
Abstract PDF(558KB) ( 557 )   
References | Related Articles | Metrics
Problem frames have been widely recognized and studied in the research domain requirements engineering.Although many research outcomes about problem frames have been obtained,how to transform requirements models (i.e.,problem diagrams) into software design artifacts and implementations smoothly is still a difficult problem to be solved.In this paper,we demonstrated how to transform problem diagrams into user scenarios,which are then used to assist behavior driven design and development of software,and we proposed a software development method which combines problem frames and behavior driven design (PFBDD).The method can help system analysts to move smoothly from requirements analysis to software design and testing,thus avoiding inconsistency problems in software development.In addition,a case study was presented to demonstrate how to apply our method into a queuing problem of a vehicle management system.The Gherkin language and Specflow tool were introduced and applied in the case study.This method plays an important role in driving problem frames into further practical applications.
Approach to C++ Code Instrumentation for Coverage Analysis with Clang/LLVM
LI Shu-fang, AN Jin-xia, LIU Yang and CHEN Liang
Computer Science. 2017, 44 (11): 191-194.  doi:10.11896/j.issn.1002-137X.2017.11.029
Abstract PDF(516KB) ( 2041 )   
References | Related Articles | Metrics
In recent years,many safety-critical software system have been running under domestic Linux operating systems,most of which are developed with C++.As C++ is expanding to new revision of the C++ ISO/IEC stan-dards,the existing source code instrumentation tools for coverage analysis cannot satisfy user’s requirements.In this paper,an approach to C++ source code instrumentation for coverage analysis with Clang/LLVM was proposed.Employing the Clang/LLVM library in open source community,which provides C++ source code parsing and manipulation functions,a framework of C++ coverage analysis for statement coverage,branch coverage and MC/DC coverage was constructed.Coverage information was collected in real-time,and was analyzed and exported to reports.Case study shows that the proposed approach is very convenient and practical in real world software testing activities.
Non-deadlock Concurrency Fault Localization Approach Based on Adequate Test Criteria
CHEN Cheng, ZHENG Zheng, WANG Hao-qin and QIAO Yu
Computer Science. 2017, 44 (11): 195-201.  doi:10.11896/j.issn.1002-137X.2017.11.030
Abstract PDF(969KB) ( 447 )   
References | Related Articles | Metrics
Concurrency programs are difficult to debug due to their congenital nondeterminism.Fault localization approaches may alleviate such situation.However,their performance heavily depends on the debug information.To allevia-te this problem,we proposed a non-deadlock concurrency fault localization approach based on adequate test criteria.Our approach consists of three parts:a prediction of the conditions meeting an adequate test criteria,a scheme of a test plan satisfied with the criteria,and a fault localization analysis using the debug information.Besides,a prototype called ConFinder based on this approach was implemented in C#.Moreover,experiments on programs containing real concurrency bugs were carried out.The result shows that our approach can localize non-deadlock concurrency bugs effectively and has stable performance.
Research on Recommendation of Concurrency Bug Testing Tools Based on Ontology
ZHENG Wei, HUANG Yue-ming, WU Xiao-xue, FENG Chen and LIN Jun
Computer Science. 2017, 44 (11): 202-206.  doi:10.11896/j.issn.1002-137X.2017.11.031
Abstract PDF(621KB) ( 425 )   
References | Related Articles | Metrics
With the concern of concurrent system testing increasing recently,new and different test tools are constantly coming into the market.For software test engineer,the ability of selecting test tools correctly has become the determining factors of testing efficiency.Because concurrent system test tools are not as well-known as those traditional test tools,we designed a recommendation approach based on ontology to promote concurrent system test tools for users.The recommendation approach recommends concurrent system test tools through three modules to improve test efficiency:the types of concurrency bugs,the features of the program to be tested and the requirement of users.
Behavioral Model Construction Method for Mobile Applications Based on Smali Code
YU Yong and GUO Qian
Computer Science. 2017, 44 (11): 207-220.  doi:10.11896/j.issn.1002-137X.2017.11.032
Abstract PDF(2304KB) ( 631 )   
References | Related Articles | Metrics
With the rapid growth in the number of mobile applications and the speed requirement of mobile application development,mobile application testing faces big challenge.In this case,many applications have been put into the market without adequate testing,which may contain bugs and impact the user’s experience.Model-driven testing method is one of the most effective testing methods,which is widely used in function,performance,availability,security testing,and can greatly improve the automation of testing.The biggest difference between the model-driven testing methods for mobile applications and traditional applications is model building method,because mobile applications are event-driven.In this paper,we proposed a method to build mobile application behavior state machine model.First,the intermediate code of the application is obtained through reverse engineering method.Then intermediate code is used to generate event table by static and dynamic methods.At last,the model is built by scanning the screen in the system layer to discover new states and define all states appeared.Experimental results show that our method can effectively and accurately build mobile application behavior model and solve some of the issues which exist in the mobile application model building.
Test Case Retrieval Method for Different Data Model
LI Yuan-yuan, HUANG Song, HUI Zhan-wei and LI Liu-yi
Computer Science. 2017, 44 (11): 221-225.  doi:10.11896/j.issn.1002-137X.2017.11.033
Abstract PDF(750KB) ( 465 )   
References | Related Articles | Metrics
The core of software testing is to design and execute test cases.In order to improve the efficiency of software testing,retrieve and share the test cases in different departments,this paper proposed the retrieval method of test case for different data model.Firstly,the test case is facetedly classified and descripted in this method,and the test case facet classification model and instantiates are proposed to provide the clues for the test case retrieval.Secondly,the test case index tree is proposed based on the instantiation of test case facet classification model,and the test cases are organized into a structure which is beneficial to the user’s retrieve.Lastly,at the time of test case retrieval for different data mo-del,through the matching of query tree and the test case index tree and keyword retrieval,it retrieves the test cases that meet the requirement from different heterogeneous databases.The experiment proves the feasibility and effectiveness of the retrieval method proposed in this paper in the practical application,which lays the foundation of test case sharing and reuse,and improves the efficiency of software testing.
Defects Detection Based on Mining Function Call Sequence Patterns
CUI Zhan-qi, MU Yong-min, ZHANG Zhi-hua and WANG Wei-guang
Computer Science. 2017, 44 (11): 226-231.  doi:10.11896/j.issn.1002-137X.2017.11.034
Abstract PDF(870KB) ( 671 )   
References | Related Articles | Metrics
Large scale programs usually imply a large number of programming rules.However,if programmers violate those rules in the process of programming,it is possible to cause software defects.The function call rule is one kind of the typical implicit rules in programs.Previous work on mining function rules handle function calls in the body of a function definition as an itemset,and the constraints implied in function call sequences are not utilized,which can lead to high false positive rates.If the function call sequence information is exploited in the process of mining rules,it will effectively improve the accuracy of mining defects.This paper proposed a defect detection approach based on mining function call sequence patterns.In the approach,the suspected defects which violate function call sequence patterns are detected automatically,and the defects with high suspicious degrees are reported.Based on this approach,experiments were carried out in a group of open source projects.The expriment results show that this approach can effectively find defects which violate function call sequence patterns in programs,and reduce false positives.As a result,the overhead of veri-fying suspicious defects are also reduced.
Model-driven Testing for Mobile Applications
FENG Gu and LI Ni-ge
Computer Science. 2017, 44 (11): 232-239.  doi:10.11896/j.issn.1002-137X.2017.11.035
Abstract PDF(1444KB) ( 504 )   
References | Related Articles | Metrics
In mobile Internet,the intelligent mobile terminal and mobile applications are widely used in all fields.At the same time,the diversity and heterogeneity of the mobile terminal hardware and platform have caused redundant work in mobile application development and testing.Cross platform issues arise and it becomes a hot area for study and practice in academic and industry.This paper proposed model driven testing for mobile applicaitons.Firstly,the method uses UML state machine to describe the behaviour of the application.Secondly,based on the behaviour model,the method generates platform independent test cases automatically.Lastly,the method maps the test cases which are unrelated to platform to multiple platforms and generates executable test cases.This paper chose a power application as an example and realized the automation test in both IOS and Android.The validity of the model driven testing method was verified in solving the cross platform problem.
Program Verification of Software Model’s Algebraic Properties
ZHAO Hui-qun and HUANG Yu-han
Computer Science. 2017, 44 (11): 240-245.  doi:10.11896/j.issn.1002-137X.2017.11.036
Abstract PDF(798KB) ( 481 )   
References | Related Articles | Metrics
Software model algebra is introduced to model software architecture by process algebra.It explains components as variables,abstracts connector as algebra operation,and builds the algebraic model of software architecture according to software features.Based on the discussion about algebraic model,we got a series of algebraic properties to guide software evolution.However,these studies have only theoretically proved the algebraic properties of the mode,no program can actually prove the correctness of these algebraic properties,and they did not give the property applications,making them lack maneuverability.This paper researched the algebraic properties of algorithms,did program verification,further enriched the software model,and also made it possible to change software evolution from theory into practical applications.
Windows Security Mechanisms Simulation and Sandbox System Implementation Based on Wine
DIAO Ming-zhi, ZHOU Yuan, LI Zhou-jun and ZHAO Yu-fei
Computer Science. 2017, 44 (11): 246-252.  doi:10.11896/j.issn.1002-137X.2017.11.037
Abstract PDF(1194KB) ( 1155 )   
References | Related Articles | Metrics
We simulated two Windows security mechanisms,adress space layout randomization (ASLR) and user account control (UAC) Virtualization,based on open source software Wine.The two mechanisms make the Wine’s environment closer to the real operating system and safer.Based on the two security mechanisms,we further presented a relatively real sandbox system,which employs the wineserver mechanism and utilizes the .wine directory of Wine as the running environment for samples to detect the dynamic behavior.The experimental results show that the proposed sandbox system presents the basic characteristics of ASLR and UAC Virtualization.Compared with other sandboxes,our proposed sandbox system can not only effectively detect behaviors of unknown samples,but also has features like low resource utilization,good isolation and fast status rollback,which make it meet the requirements of batch deployment and operation.
Certificateless Threshold Decryption Scheme for Data Security of Recommendation System
ZHU Jun, YUAN Xiao-feng, GOU Zhi-nan and YANG Yi
Computer Science. 2017, 44 (11): 253-263.  doi:10.11896/j.issn.1002-137X.2017.11.038
Abstract PDF(1781KB) ( 597 )   
References | Related Articles | Metrics
Recommendation system is an effective way to solve the problem of information overloading and meanwhile satisfy user’s personalized needs.Nevertheless the data security issues involved from the recommendation procedure are definitely hindering the healthy development of recommendation system.In distributed recommendation systems,thre-shold decryption is one of the useful methods to resist security attack and protect user’s privacy.Regarding aforementioned techniques,we studied threshold decryption in the context of certificateless public key cryptography,aiming to avoid costly management of certificate in public key infrastructure and settle the matter of key escrow in identity-based cryptography.This paper introduced a new construction for certificateless threshold decryption scheme and the corresponding security model.The scheme was proved secure against chosen-ciphertext attack in the random oracle model and the security proof was presented under the condition of hard computation of a problem in relation to bilinear diffie-hellman problem.Compared with other existing schemes,our scheme has lower computational overhead,faster transmission rate and shorter master secret key and public key.The scheme can not only improve the efficiency of recommendation system but also effectively ensure the safety of user’s privacy.
Sub-network Preclusion in (n,k)-bubble-sort Networks
YANG Yu-xing and QIU Ya-na
Computer Science. 2017, 44 (11): 264-267.  doi:10.11896/j.issn.1002-137X.2017.11.039
Abstract PDF(540KB) ( 508 )   
References | Related Articles | Metrics
In the real parallel computer systems,faults of processors and links are inevitable,and the fault tolerance abi-lity of a system can be evaluated by the performance of its interconnection network.In order to measure the fault tole-rance abilities of the parallel computers which take (n,k)-bubble-sort networks as underlying topologies,combining the hierarchical structures and sub-network partitions’ characters of (n,k)-bubble-sort networks,an algorithm of the problem that destroy all the (n-m,k-m)-bubble-sort networks in the (n,k)-bubble-sort network under the node fault model and the link fault model was proposed,and the optimal attacking nodes set and the optimal attacking links set were obtained.According to the algorithm,to destroy all the (n-m,k-m)-bubble-sort networks in the (n,k)-bubble-sort network,at least Cmnm!nodes or Cmnm!links ought to be faulty under the node fault model and the link fault mo-del,respectively,where 2≤k≤n-2 and m≤k-1.
P2P Based Massive Scalable Remote Sensing Data Distribution with Access Control
YANG Xin-lei, HE Qian, CAO Li and WANG Shi-cheng
Computer Science. 2017, 44 (11): 268-272.  doi:10.11896/j.issn.1002-137X.2017.11.040
Abstract PDF(965KB) ( 505 )   
References | Related Articles | Metrics
With the increasing of remote sensing data,massive scalable remote sensing data distribution may cause big pressure for the distribution center server.Fully utilizing the network resources of joining nodes,a P2P based massive scalable remote sensing data distribution system was proposed and realized.The remote sensing data distribution system consists of remote sensing data management platform and remote sensing data client,where the remote sensing data mangement is composed of the sharing distribution platform web site,the cloud storage,the seed resource server and the tracker,and these remote sensing clients and the seed resource server construct a P2P network.The P2P communication protocol including sharing pieces,piece selection,and tracker communication was designed.The realized system can support uploading data and begining seed automatically with user access control.The users download the remote sensing data based on their user ranks and share and fasten data pieces downloading based on the Bittorrent like protocol.The experiment results show that the realized massive scalable remote sensing data distribution system is fully functional,has excellent concurrent performance while multiple nodes downloading,and satisfies the requirements of massive scalable remote sensing data distribution.
Chosen Plaintext Attack on Image Encryption Algorithm Based on Bit Scrambling and Hyperchaos
ZHU Shu-qin, WANG Wen-hong and SUN Zhong-gui
Computer Science. 2017, 44 (11): 273-278.  doi:10.11896/j.issn.1002-137X.2017.11.041
Abstract PDF(986KB) ( 469 )   
References | Related Articles | Metrics
Recently,an image encryption algorithm based on bit scrambling and hyperchaos was proposed,whose main idea is as follows.First,a chaotic sequence is used to scramble pixel values of plaintext image.Then,according to the size of two adjacent elements in a random sequence,internal bit of each pixel is scrambled.In the end,the final ciphertext image is obtained by the diffusion and confusion of the scrambling sequence,and chaotic sequence,and plain image information can be well hidden.In this paper,the security of the above encryption algorithm were analyzed.It is found that the security of the algorithm depends entirely on three chaotic sequences by choosing plaintext attack.The three chaotic random sequences in the original algorithm were obtained,thus the plaintext image was restored.Theoretical analysis and experimental results verify the feasibility of the chosen plaintext attack strategy.At the same time,the improved method was given.In the improved algorithm,the initial value of the chaotic system is related to the SHA-256 hash value of the plaintext image,so that the key stream is related to the plaintext image.Then the algorithm is resis-tant to the attack of chosen plaintext.
Path Identification Based Delaying Restart Algorithm for CDCL SAT Solver
CHEN Qing-shan, XU Yang, WU Guan-feng and HE Xing-xing
Computer Science. 2017, 44 (11): 279-283.  doi:10.11896/j.issn.1002-137X.2017.11.042
Abstract PDF(698KB) ( 548 )   
References | Related Articles | Metrics
An appropriate restart is helpful for a solver to jump out of the local optimization,but more frequent restarts will significantly reduce the efficiency.To address the arbitrariness of triggering conditions for the restart of CDCL solver,the delaying restart algorithm based on path identification was proposed in this paper.Specifically,the Luby sequence is utilized to trigger the delaying restart decision,which converts current path and the searched paths to vector space models (VSMs) such that the similarity of VSMs is calculated to judge whether the current search process will get into the repetitive search space.Once the underlying similarity reaches a given threshold,the restart is triggered,otherwise it will be delayed.SAT international testing example and two state-of-the-art solvers were adopted for comparison purpose.The experimental results show that the proposed algorithm can not only effectively avoid the repetitive search space,but also obviously improve the solving efficiency.
h-ACOR:An ACOR Algorithm with Heuristic Information for Neural Network Training
ZHAO Zhang-ming, FENG Jing, SHI En and SHU Xiao-cun
Computer Science. 2017, 44 (11): 284-288.  doi:10.11896/j.issn.1002-137X.2017.11.043
Abstract PDF(834KB) ( 596 )   
References | Related Articles | Metrics
The ACOR algorithm is an ant colony optimization(ACO) algorithm for real-valued optimization.The ACOR can be used for training neural network.Unlike most of the conventional ACO algorithms,ACOR does not consider heuristic information when training neural networks.So in this work,a new algorithm named h-ACOR that incorporates the heuristic information into the framework of ACOR was proposed for neural network training.The heuristic information in h-ACOR is a gradient vector,which is obtained by computing the partial derivative of error term of the neural network with respect to weight vector.Using 10-fold cross-validation method,h-ACOR is applied to train neural networks for pattern classification problems of zoo,iris and tic-tac-toe in UCI datasets.Compared with ACOR,h-ACOR can reduce classification errors while speeding up the convergence process,with the average training generations of h-ACOR being nearly 1/2 of that of ACOR.After completely training by h-ACOR,the classification accuracy about zoo,iris and tic-tac-toe are respectively 91.1%,93.3% and 95.6%,which have better performance than that of ACOR’s 83.1%,88.7% and 91.9%.
Similarity Measure for Text Classification Based on Feature Subjection Degree
CHI Yun-xian, ZHAO Shu-liang, LUO Yan, ZHAO Jun-peng, GAO Lin and LI Chao
Computer Science. 2017, 44 (11): 289-296.  doi:10.11896/j.issn.1002-137X.2017.11.044
Abstract PDF(1009KB) ( 547 )   
References | Related Articles | Metrics
It is a fashionable method to do text classification based on similarity.Algorithm similarity measure for text classification based on feature subjection degree (SMTCFSD) aims at measuring similarity of documents through subjection relationship between feature words and documents.Algorithm SMTCFSD divides words into total subjection word sets,partial subjection word sets and none subjection word sets based on the subjection relationship,and defines subjection function based on three subjection word sets.Total subjection word sets subject to two documents,and subjection degree will decrease when the differences between two total subjection words increase.The words that only belong to one of the two documents are subsumed into partial subjection word sets,in which subjection degree is a definite value.Subjection degree of none subjection word sets is zero,because the words subject to neither of two documents.Total subjection relationship is more important than partial subjection relationship for similarity measure.Due to word sets of documents in the same category is similar to each other,while the ones in different categories have great distinction,classification accuracy will be promoted obviously based on different values of feature words,which are decided by subjection degree.Algorithm SMTCFSD is superior to the widely used similarity measure methods through experimental results on data sets from Reuters-21578 and 20-Newsgroups.
Medical Image Registration Based on Self-adaptive DE Algorithm and Powell Algorithm
LIU Zhe, SONG Yu-qing and WANG Dong-dong
Computer Science. 2017, 44 (11): 297-300.  doi:10.11896/j.issn.1002-137X.2017.11.045
Abstract PDF(587KB) ( 476 )   
References | Related Articles | Metrics
Image registration is a key technology in medical image processing.This paper proposed a new multi-resolution medical image registration method based on self-adaptive difference algorithm (DE) and Powell algorithm.It can not only overcome the shortcomings of Powell algorithm depending on the initial,but also can reduce the possibility of getting into local extreme value.Firstly,the source image is processed by multi resolution,and the three layer image including the source image is obtained.Secondly,the adaptive DE algorithm is used to search the global transform parameters on the low resolution images.The transformation parameters are obtained as the initial points of the Powell algorithm.Finally,the Powell algorithm is used for registration in both high resolution images and source images.Compared with traditional experiment,this method has higher precision and can effectively avoid local convergence problem.
Research of Document Image Super Resolution Algorithm Based on Directional Bilateral Total Variation Regularization
LIU Xiao-yuan, YI Yang, YANG Lei and WANG Bin
Computer Science. 2017, 44 (11): 301-304.  doi:10.11896/j.issn.1002-137X.2017.11.046
Abstract PDF(629KB) ( 616 )   
References | Related Articles | Metrics
For super resolution image reconstruction problem in the traditional document,based on bilateral total variation regularization super-resolution algorithm,a document image super resolution algorithm based on improved BTV was proposed.By introducing a new regularization term,which is the direction of the stroke width,the algorithm is adaptive to the local width and the local direction of the character strokes.Then,the proposed algorithm can make the local stroke width of the output image close to the local stroke direction by analyzing the low resolution image and its interpolation.This information is compressed into the directional total variation regularization term based on the width of stroke.By linear combination of minimization regularization and data fidelity term,the high-resolution images are reconstructed.Compared with the related document image super resolution method,the proposed method has been improved in the visual image quality and character recognition accuracy.
Robust Multi-lane Detection Algorithm
XUAN Han-yu, LIU Hong-zhe, YUAN Jia-zheng, LI Qing and NIU Xiao-ning
Computer Science. 2017, 44 (11): 305-313.  doi:10.11896/j.issn.1002-137X.2017.11.047
Abstract PDF(1511KB) ( 1214 )   
References | Related Articles | Metrics
Machine vision is widely used in the field of intelligent vehicles,lane line detection has become a research hotspot in the field of artificial intelligence.In order to obtain a more robust lane-line detection effect,a lane-line feature filter based on multi-condition constraint was proposed,and a new clustering algorithm for lane-line feature was proposed.By using Kalman filter,the position and width of the lane line based on the perspective projection linear relationship are used to set the adaptive dynamic ROI.In addition,the constraint condition is obtained by using the characteristics of the lane line to obtain a more stable detection result.The real-time detection results in the real road environment show that the algorithm is robust and real-time,and has a very stable detection effect.
Co-saliency Detection via Superpixel Matching
ZHANG Zhao-feng, WU Ze-min, JIANG Qing-zhu, DU Lin and HU Lei
Computer Science. 2017, 44 (11): 314-319.  doi:10.11896/j.issn.1002-137X.2017.11.048
Abstract PDF(941KB) ( 501 )   
References | Related Articles | Metrics
To effectively address the issue of multi-scene co-saliency detection,we proposed a novel model based on superpixel matching and cellular automata.First of all,we introduced an adjacent superpixel sets matching algorithm based on Hausdorff distance to achieve exact matching between image supperpixels.Comparing to the traditional superpixel matching algorithm,the new algorithm greatly improves the matching accuracy.In addition,we further proposed the 2-layer cellular automata via intra image and inter images to carry out the significant propagation of multiple images,thus exploit the intrinsic relevance of similar regions through interactions with neighbors in multi-scene.Experimental results demonstrate that our model outperforms state-of-the-art methods.Furthermore,the proposed methods is efficient and robust.