计算机科学 ›› 2021, Vol. 48 ›› Issue (7): 213-218.doi: 10.11896/jsjkx.200800127

• 计算机图形学&多媒体 • 上一篇    下一篇

结合乐高滤波器和SSD的低光照图像融合检测方法

李琳1, 刘学亮1, 赵烨1, 纪平2   

  1. 1 合肥工业大学计算机与信息学院 合肥230601
    2 合肥学院电子信息与电气工程系 合肥230601
  • 收稿日期:2020-08-19 修回日期:2020-10-02 出版日期:2021-07-15 发布日期:2021-07-02
  • 通讯作者: 刘学亮(Liuxueliang1982@gmail.com)
  • 基金资助:
    科技部重点研发计划(2018AAA0102002);国家自然科学基金(61976076,61632007,61932009,61806066);安徽省高校自然科学研究项目(KJ2018A0545)

Low Light Image Fusion Detection Method Based on Lego Filter and SSD

LI Lin1, LIU Xue-liang1, ZHAO Ye1, JI Ping2   

  1. 1 School of Computer and Information Engineering,Hefei University of Technology,Hefei 230601,China
    2 Department of Electronic Information and Electrical Engineering,Hefei University,Hefei 230601,China
  • Received:2020-08-19 Revised:2020-10-02 Online:2021-07-15 Published:2021-07-02
  • About author:LI Lin,born in 1994,postgraduate.Her main research interests include object detection and computer vision.(804082803@qq.com)
    LIU Xue-liang,born in 1981,Ph.D,associate professor,is a member of China Computer Federation.His main research interests include multimedia information retrieval and so on.
  • Supported by:
    Key R&D plan of the Ministry of Science and Technology(2018AAA0102002),National Natural Science Foundation of China(61976076,61632007,61932009,61806066) and Natural Science Research Projects of Universities in Anhui Province(KJ2018A0545).

摘要: 针对低光照图像背景环境复杂导致目标检测易产生误检、漏检现象,提出了一种基于SSD目标检测的改进低光照图像精度和速度的方法。该方法先对低光照图像进行增强处理,然后将处理后的低光照图像和增强图像分别输入到融入乐高滤波器的SSD网络结构中进行训练检测,通过得到的两种检测模型对处理后的数据集进行检测,最后融合检测结果候选框中的不重复框,筛选候选框中的重复框,标记出正确位置的目标,从而提升对低光照图像检测的精度。在网络结构不同位置融入乐高滤波器,模型参数量分别减少8.9%和29.5%,浮点运算次数下降6.8%和34.9%,检测框融合处理后检测精度得到了3%~7%的提升。该方法更符合实际应用,有效提升了低光照图像的检测速度和精度,扩大了目标检测的应用范围。

关键词: SSD算法, 低光照图像, 乐高滤波器, 目标检测, 融合

Abstract: Aiming at the problem that target detection is prone to misdetection or missing detection due to the complex background environment of low light image,this paper proposes a method to improve the accuracy and speed of low light image based on SSD object detection.Firstly,the low light image is enhanced,the processed enhanced image and the original low light image are respectively input into the SSD network structure with Lego filter for training detection.The two detection models are used to train and detect the enhanced data set,and a series of candidate frames are obtained.Finally,the non-repeated frames in the candidate frames are fused to mark the target in the correct position,so as to improve the detection accuracy of low light image.At the same time,Lego filter is integrated into the network structure to reduce the model parameters of network training,so as to improve the detection speed.Experimental results show that,when Lego filter is integrated in different positions of the network structure,the parameters of the model reduce by 8.9% and 29.5%,and the numbers of floating-point operations reduce by 6.8% and 34.9%.After fusion processing,the detection accuracy improves by 3% ~7%.This method is more suitable for practical application,effectively improves the detection speed and accuracy of low light image,and expands the application range of object detection.

Key words: Fusio, Lego filter, Low light image, Object detection, Single Shot MultiBox Detector algorithm

中图分类号: 

  • TP391
[1]ANDREOPOULOS A,TSOTSOS J K.50 years of object recognition:Directions forward[J].Computer Vision and Image Understanding,2013,117(8):827-891.
[2]LORE K G,AKINTAYO A,SARKAR S.LLnet:a deep autoencoder approach to natural low-light image enhancement[J].Pattern Recognition,2017,61:650-662.
[3]WEI C,WANG W,YANG W,et al.Deep retinex decomposition for low-light enhancement[J].arXiv:1808.04560, 2018.
[4]ZHANG Y H,ZHANG J W,GUO X J.Kindling the Darkness:A Practical Low-light Image Enhancer[C]//Proceedings of the 27th ACM International Conference on Multimedia.2019:1632-1640.
[5]ZEILER M D,FERGUS R.Visualizing and Understanding Convolutional Networks[C]//European Conference on Computer Vision.Springer,2014:818-833.
[6]KRIZHEVSKY A,SUTSKEVER I,HINTON G E.Imagenetclassification with deep convolutional neural networks[C]//Advances in Neural Information Processing Systems.2012:1097-1105.
[7]REDMON J,DIVVALA S,GIRSHICK R.You Only LookOnce:Unified,Real Time Object Detection[C]//Computer Vision & Pattern Recognition.IEEE,2016:13.
[8]LIU W,DRAGOMIR A,DUMITRU E.SSD:single shot multibox detector[C]//Proc of Computer Vision and Pattern Recognition.2016:21-37.
[9]GIRSHICK R,DONAHUE J,DARRELL T,et al.Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.2014:580-587.
[10]EVERINGHAM M,GOOL L V,WILLIAMS C K I,et al.The Pascal Visual Object Classes (VOC) Challenge[J].International Journal of Computer Vision,2010,88(2):303-338.
[11]LIN T,MAIRE M,BELONGIE S,et al.Microsoft coco:Co-mmon objects in context[C]//European Conference on Compu-ter Vision.Cham:Springer,2014:740-755.
[12]ZOU Z X,SHI Z W,GUO Y H.Object Detection in 20 Years:A Survey[J].arXiv:1905.05055v2,2019.
[13]ZHANG Q,YUAN G,XIAO C,et al.High-quality exposurecorrection of underexposed photos[C]//Proceedings of the 26th ACM international conference on Multimedia.2018:582-590.
[14]RASMUS R,GUILLAUMIN M,GOOL L V.Non-maximumSuppression for Object Detection by Passing Messages Between Windows[C]//Asian Conference on Computer Vision.Cham:Springer,2014.
[15]BAI C,HUANG L,CHEN J N,et al.Optimization of deep con-volutional neural network for large scale image classification[J].Journal of Software,2018,29(4):1029-1038.
[16]LIN T Y,DOLLÁR P,GIRSHICK R,et al.Feature pyramidnetworks for object detection[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.2017:2117-2125.
[17]YANG Z,WANG Y,LIU C,et al.Legonet:Efficient convolu-tional neural networks with lego filters[C]//International Conference on Machine Learning.PMLR,2019:7005-7014.
[18]NAVANEETH B,BHARAT S,RAMA C,et al.Soft-nms-improving object detection with one line of code[C]//Proc of IEEE International Conference on Computer Vision.2017:5562-5570.
[19]LOH Y P,CHAN C S.Getting to know low-light images with the exclusively dark dataset[J].Computer Vision and Image Understanding.2019:30-42.
[20]FU X,ZENG D,HUANG Y,et al.A fusion-based enhancingmethod for weakly illuminated images[J].Signal Processing,2016,129:82-96.
[21]GUO X,LI Y,LING H.Lime:Low-light image enhancement via illumination map estimation[J].IEEE Trans on Image Proces-sing,2017,26:982-993.
[22]MITTAL A,SOUNDARARAJAN R,BOVIK A.Making a“Completely Blind” Image Quality Analyzer[J].IEEE Signal Processing Letters,2013,20(3):209-212.
[23]YUE G,HOU C,ZHOU T.Blind quality assessment of tone-mapped images considering colorfulness,naturalness,and structure[J].IEEE Transactions on Industrial Electronics,2019,66(5):3784-3793.
[24]ZHAI G,WU X,YANG X,et al.A Psychovisual Quality Metric in Free-Energy Principle[J].IEEE Transactions on Image Processing,2012,21(1):41-52.
[25]GU K,LIN W,ZHAI G,et al.No-reference quality metric of contrast-distorted images based on information maximization[J].IEEE Transactions on Cybernetics,2017,47(12):4559-4565.
[1] 吴子仪, 李邵梅, 姜梦函, 张建朋.
基于自注意力模型的本体对齐方法
Ontology Alignment Method Based on Self-attention
计算机科学, 2022, 49(9): 215-220. https://doi.org/10.11896/jsjkx.210700190
[2] 曹晓雯, 梁美玉, 鲁康康.
基于细粒度语义推理的跨媒体双路对抗哈希学习模型
Fine-grained Semantic Reasoning Based Cross-media Dual-way Adversarial Hashing Learning Model
计算机科学, 2022, 49(9): 123-131. https://doi.org/10.11896/jsjkx.220600011
[3] 周乐员, 张剑华, 袁甜甜, 陈胜勇.
多层注意力机制融合的序列到序列中国连续手语识别和翻译
Sequence-to-Sequence Chinese Continuous Sign Language Recognition and Translation with Multi- layer Attention Mechanism Fusion
计算机科学, 2022, 49(9): 155-161. https://doi.org/10.11896/jsjkx.210800026
[4] 汪鸣, 彭舰, 黄飞虎.
基于多时间尺度时空图网络的交通流量预测模型
Multi-time Scale Spatial-Temporal Graph Neural Network for Traffic Flow Prediction
计算机科学, 2022, 49(8): 40-48. https://doi.org/10.11896/jsjkx.220100188
[5] 秦琪琦, 张月琴, 王润泽, 张泽华.
基于知识图谱的层次粒化推荐方法
Hierarchical Granulation Recommendation Method Based on Knowledge Graph
计算机科学, 2022, 49(8): 64-69. https://doi.org/10.11896/jsjkx.210600111
[6] 魏恺轩, 付莹.
基于重参数化多尺度融合网络的高效极暗光原始图像降噪
Re-parameterized Multi-scale Fusion Network for Efficient Extreme Low-light Raw Denoising
计算机科学, 2022, 49(8): 120-126. https://doi.org/10.11896/jsjkx.220200179
[7] 刘冬梅, 徐洋, 吴泽彬, 刘倩, 宋斌, 韦志辉.
基于边框距离度量的增量目标检测方法
Incremental Object Detection Method Based on Border Distance Measurement
计算机科学, 2022, 49(8): 136-142. https://doi.org/10.11896/jsjkx.220100132
[8] 王灿, 刘永坚, 解庆, 马艳春.
基于软标签和样本权重优化的Anchor Free目标检测算法
Anchor Free Object Detection Algorithm Based on Soft Label and Sample Weight Optimization
计算机科学, 2022, 49(8): 157-164. https://doi.org/10.11896/jsjkx.210600240
[9] 沈祥培, 丁彦蕊.
多检测器融合的深度相关滤波视频多目标跟踪算法
Multi-detector Fusion-based Depth Correlation Filtering Video Multi-target Tracking Algorithm
计算机科学, 2022, 49(8): 184-190. https://doi.org/10.11896/jsjkx.210600004
[10] 闫佳丹, 贾彩燕.
基于双图神经网络信息融合的文本分类方法
Text Classification Method Based on Information Fusion of Dual-graph Neural Network
计算机科学, 2022, 49(8): 230-236. https://doi.org/10.11896/jsjkx.210600042
[11] 陈明鑫, 张钧波, 李天瑞.
联邦学习攻防研究综述
Survey on Attacks and Defenses in Federated Learning
计算机科学, 2022, 49(7): 310-323. https://doi.org/10.11896/jsjkx.211000079
[12] 张源, 康乐, 宫朝辉, 张志鸿.
基于Bi-LSTM的期货市场关联交易行为检测方法
Related Transaction Behavior Detection in Futures Market Based on Bi-LSTM
计算机科学, 2022, 49(7): 31-39. https://doi.org/10.11896/jsjkx.210400304
[13] 张颖涛, 张杰, 张睿, 张文强.
全局信息引导的真实图像风格迁移
Photorealistic Style Transfer Guided by Global Information
计算机科学, 2022, 49(7): 100-105. https://doi.org/10.11896/jsjkx.210600036
[14] 程成, 降爱莲.
基于多路径特征提取的实时语义分割方法
Real-time Semantic Segmentation Method Based on Multi-path Feature Extraction
计算机科学, 2022, 49(7): 120-126. https://doi.org/10.11896/jsjkx.210500157
[15] 郁舒昊, 周辉, 叶春杨, 王太正.
SDFA:基于多特征融合的船舶轨迹聚类方法研究
SDFA:Study on Ship Trajectory Clustering Method Based on Multi-feature Fusion
计算机科学, 2022, 49(6A): 256-260. https://doi.org/10.11896/jsjkx.211100253
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!