基于全景鸟瞰视图的障碍物检测方法研究

计算机科学 ›› 2014, Vol. 41 ›› Issue (Z11): 78-82.

基于全景鸟瞰视图的障碍物检测方法研究

常嘉义,秦瑞,李庆,陈曦,徐坚俊

中国科学院微电子研究所北京100871;中科院微电子所昆山分所昆山215347;中国科学院微电子研究所北京100871;中科院微电子所昆山分所昆山215347;中国科学院微电子研究所北京100871;中科院微电子所昆山分所昆山215347;中科院微电子所昆山分所昆山215347;中国科学院无锡物联网研究发展中心无锡214135;中国科学院无锡物联网研究发展中心无锡214135

出版日期:2018-11-14 发布日期:2018-11-14
基金资助:
本文受汽车移动物联网总体技术研究(2012ZX03005018),智能交通创新服务系统(XDA06040300),国家自然科学基金NSFC:面向“智慧工厂”的视觉目标检测和智能决策研究(61303174),面向智慧交通的车联网高安全性内容保障技术研究与核心产品研发(2012B0910),广东省教育部产学研结合项目资助

Research of Obstacle Detection Based on Aerial Panorama Image

CHANG Jia-yi,QIN Rui,LI Qing,CHEN Xi and XU Jian-jun

Online:2018-11-14 Published:2018-11-14

摘要/Abstract

摘要： 目前,车载环视系统的障碍物检测大多采用基于特征的方法,主要针对车辆和行人等特定障碍物,不能完全消除障碍物对车辆行驶安全的威胁。采取基于运动的方法检测所有高于地面的障碍物。建立了平坦路面上的车辆运动模型,使用逆投影变换,推导出车辆运动参数、障碍物点高度和对应像素光流的关系。根据车载环视系统中相邻摄像头生成的俯视图存在重叠区域、重叠区域的障碍物特征点存在两个不同的光流矢量的特点,快速筛选出相交区域中的障碍点,并加速最优车辆运动参数的求取。最后使用运动参数生成运动补偿图像,检测图像中的所有障碍物点。实际道路测试证明,提出的算法在常规的路面环境下,能有效地标记出行驶过程中高于地面的所有障碍物。

关键词: 车载环视系统,逆投影变换,光流运动,自车运动估计

Abstract: At present,feature-based method is used to detect vehicles and pedestrians of panorama images,can not avoid all the threats of obstacles to driving.Different from the method above,we detected all the obstacles taller than the ground plane using motion-based method.First,we built the vehicle motion model,then deducd the relation-ship of vehicle motion,obstacle height and pixel optical-flow.Taking advantage of the fact that vertical views of adjacent cameras overlap each other and obstacle points in overlapping region have two different optical flows,we detected obstacles in intersection areas,fastened the solving of optimal motion parameters.At last,we detected all the obstacles using the motion compensated image.The vehicle experiments on the road demonstrate that the method proposed by us can detect all the obstacles taller than the ground effectively when vehicles are on the flat road.

Key words: Vehicle around view system,Inverse projection transformation,Optical-flow,Ego-motion estimation

常嘉义,秦瑞,李庆,陈曦,徐坚俊. 基于全景鸟瞰视图的障碍物检测方法研究[J]. 计算机科学, 2014, 41(Z11): 78-82. https://doi.org/

CHANG Jia-yi,QIN Rui,LI Qing,CHEN Xi and XU Jian-jun. Research of Obstacle Detection Based on Aerial Panorama Image[J]. Computer Science, 2014, 41(Z11): 78-82. https://doi.org/

参考文献

[1] Stein G P,Mano O,Shashua A.A robust method for computing vehicle ego-motion[C]∥Proceedings of the IEEE Intelligent Vehicles Symposium,2000(IV 2000).IEEE,2000:362-368
[2] Matthews N D,Greenway P.A Mathematical Framework forRobust Obstacle Detection Using Feature Matching[C]∥Charnley D(ed) 1st IFAC International Workshop on Intelligent Autonomous Vehicles IAV.1993:18-21
[3] Krüger W.Robust real-time ground plane motion compensation from a moving vehicle[J].Machine Vision and Applications,1999,11(4):203-212
[4] Bar-Shalom Y,Li X R,Kirubarajan T.Estimation with applications to tracking and navigation:theory algorithms and software[M].John Wiley & Sons,2004
[5] Gandhi T,Trivedi M.Motion analysis of omni-directional video streams for a mobile sentry[C]∥First ACM SIGMM international workshop on Video surveillance.ACM,2003:49-58
[6] FUJITSU.http://www.fujitsu.com/us/semiconductors/gdc/products/omni.html
[7] Gandhi T,Trivedi M.Motion based vehicle surround analysis using an omni-directional camera[C]∥Intelligent Vehicles Symposium,2004 IEEE.IEEE,2004:560-565
[8] Broggi A,Bertozzi M,Fascioli A,et al.Shape-based pedestriandetection[C]∥Proceedings of the IEEE Intelligent Vehicles Symposium.2000:215-220
[9] Shashua A,Gdalyahu Y,Hayun G.Pedestrian detection for dri-ving assistance systems:Single-frame classification and system level performance[C]∥Intelligent Vehicles Symposium,2004 IEEE.IEEE,2004:1-6
[10] Bertozzi M,Broggi A.GOLD:A parallel real-time stereo vision system for generic obstacle and lane detection[J].IEEE Transactions on Image Processing,1998,7(1):62-81
[11] 丁鑫.全景视觉泊车辅助系统研究[D].杭州:浙江大学,2010
[12] 鲁斌,秦瑞,李庆,等.车载环视拼接方法的研究[J].计算机科学,2013,40(9):293-295
[13] 赵逢.基于单目视觉的后方障碍物检测算法的研究与实现[D].沈阳:东北大学,2008
[14] 王红伟.基于单目视觉三维重建的障碍物检测算法的设计与实现[D].沈阳:东北大学,2008

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed