一种基于IPM-DVS的车道线检测算法
来源:99网
2015年4月 北京联合大学学报 Apr.2015 第29卷第2期总100期 Journal of Beijing Union University V0I.29 No.2 Sum No.100 DOI:10.16255/j.cnki.Idxbz.2015.02.008 一种基于IPM—DVS的车道线检测算法 郑永荣,袁家政,刘宏哲 (北京市信息服务工程重点实验室 北京联合大学,北京 100101) [摘 要] 为了提高在复杂情形下车道线检测的鲁棒性,提出了一种基于IPM-DVS的车道线检 测算法。首先,对视频序列中连续两帧图像进行IPM得到鸟瞰图像,对这两帧鸟瞰图像进行差值 运算。对差值图像进行Sobel算子卷积,然后将卷积结果与鸟瞰图像进行DVS,实现车道线信息的 分离,最后根据车道线特征信息进行滤波检测。通过阴影、水渍、逆光等场景对该方法进行测试, 实验结果表明:该方法能在各种复杂环境中检测出车道线,具有实时性好、鲁棒性强、正确率高的 优点,适用于无人驾驶智能车视觉导航。 [关键词] 车道线检测;IPM.DVS;智能车 [中图分类号] u 46 [文献标志码] A [文章编号] 1005—0310(2015)02—0041—06 An Algorithm of Lane Detection Based on IPM—DVS ZHENG Yong-rong,YUAN Jia-zheng,LIU Hong—zhe (Beijing Key Laboratory of Information Service Engineering,Beijing Union University,Beijing 100101,China) Abstract:To improve the robustness of lane detection under complex conditions,proposed an algorithm of lane detection based on IPM-DVS.First.get the bird’s—eye images of the continuous two original images by IPM. Then,execute a Sobel operator convolution on the image of difference operation for the bird’s—eye images.After that,separate out the lane information from the original image via DVS of the result image of Sobel operator convolution and the IPM image.Finally,detect the lanes through the lane features filtering.The proposed algorithm is tested in several situations,including in the presence of shadow effects,waterlogging,and baeklighting.The experimental results show that the method can detect the lane—marks in real time and robustness for various complex environments.The lane detection algorithm can be applied to the vision navigation for unmanned intelligent vehicles. Key words:Lane detection;IPM—DVS;Intelligent vehicle 0 引言 熟,刘欣等人提出的一种基于视觉长距离预测的车 道线检测方法 已经在无人驾驶智能车平台得到 近年来,无人驾驶智能车的研究已成为人工智 验证。针对半结构化道路(城市道路)环境 ’ , 能和计算机视觉的研究热点,而车道线检测是无人 陈培等人提出了一种复杂环境中的车道线检测方 驾驶智能车视觉导航的重要组成部分 。结构化 法¨ ,该方法对复杂环境(如阴影、雨天、夜晚、山的 道路(高速公路)上的车道线检测算法 比较成 影响)有一定鲁棒性,但是还没有解决逆光带来的 [收稿日期] 20l5一O3一O5 [基金项目] 北京市教育委员会科技发展计划面上项目(SQKM201411417004),北京市教育委员会创新团队项目 (IDHT20140508),北京联合大学人才强校计划人才资助项目(BPHR2014E02)。 [通信作者] 刘宏哲,E—mail:liuhongzhe@buu.edu.12n 42 北京联合大学学报 2015年4月 影响,并且其采用的是非IPM检测方法 ],不能 够为智能车提供精确视觉导航轨迹。车道线检测 关键的步骤是车道线特征提取 ,常用的特征有颜 色特征 、梯度特征[]6-1s]、线性特征 驯和结构 特征 。在逆透视图像中车道线的结构特征比较 : 3] 突出,如车道线相互平行、车道线的宽度固定等,文 献[21]就是利用逆透视图像中车道线的结构特征 进行距离变换实现车道线检测的,该方法在直道上 检测的正确率较高,但在弯道处的检测还不够理 想,因此比较适合直道上的视觉导航。 现有的车道线检测方法的大体思路都是先对 原始图像进行灰度化和二值化的预处理 。。 ,然后 根据车道线的特征进行检测。这样的检测方法会 带来两个问题:一是丢失了原始图像的大量有效信 息,如颜色特征、纹理细节等;二是在二值化处理的 过程中如果采用固定阈值,则无法适应各种环境变 化,而采用自适应阈值算法,则会增加运算开销,使 车道线检测算法很难达到实时性。为了避免上述 两个问题的出现,本文提出了一种基于IPM.DVS 的车道线检测算法,该算法充分利用彩色图像的丰 富信息,通过逆透视变换差值分离(Inverse Perspective Mapping—Difference Vable Separation, IPM.DVS)的方法将车道线信息从彩色图像中分离 出来。1PM—DVS方法就是对彩色的逆透视图像进 行差值分离,其对线状信息具有较好的分离能力, 结合车道线的线性特征,适于车道线信息的分离。 分离出来的车道线信息与路面信息具有较大的差 异性,存在较大的灰度梯度,可采用固定阈值进行 二值化处理,提高算法的实时性,从而克服了现有 算法采用自适应阈值的缺点。 1 逆透视变换差值分离(IPM—DVS) 原理 1.1逆透视变换(IPM) 逆透视变换(Inverse Perspective Mapping,IPM) 是指将透视图像转换成俯视图像的过程,如图1所 示 图1逆透视变换过程 Fig.1 The process of the IPM 将透视图像转换成俯视图像时需要一个转换 ㈩ 日 H 『0(6, Dst( ,Y )=1 SrcA( ,Yj)一SrcB( ,Yj)1。(7) 北京联合大学学报 2015年4月 [3]Zhao Hongying,Kim O,Won J S,et a1.Lane detection and tracking based on annealed particle filter[J].Internationa1 Journal of Control,Automation,and Systems,2014,6(12):1303—1312. [4] Hur J,Kang S N,Seo S W.Multi-lane detection in urban driving environments using conditional random fields[c]//2013 IEEE Intelligent Vehicles Symposium(IV)June 23—26,2013,Gold Coast,Australia:1297—1302. [5] Kazuyuki K,Kajipo W,Tomoyuki O,et a1.A lane detection algorithm for pe ̄onal vehicles[J].Electircal Engineering in Japan,2011,4(177):23—32. [6] Wu C F,Lin C J,Lin H Y,et a1.Adjacent lane detection and lateral vehicle distance measurement using vision.based neuro— fuzzy approaches[J].Journal of Applied Research and Technology,2013(11):251—258. [7] Liu Xin,Xu Xin,Dai Bin.Vision-based long distance lane perception and front vehicle location for full autonomous vehic1es on highway roads[J].Cent South Univ,2012(19):1454—1465. [8] Choi H C,Park J M,Chou W S.Vision—based fusion of robust lane tracking and forward vehicle detection in a real driving environment[J].International Journal of Automotive Technology,2012,4(13):653—669. [9] Wang Huan,Shao Sulin.Lane markers detection based on consecutive threshold segmentation[J].Advanced Materials Research Vols,201I(317—319):881—885. [1O] Schmidt K,Matthias Beggiato,Karl Heinz Hoffmann,et a1.A mathematical model for predicting lane changes using the steering wheel angle[J].Journal of Safety Research,2014(49):85—9O. [11] Li Shasha.The research of the lane detection algorithm base on vision sensor[J].Research Journal of Applied Sciences, Engineering and Technology,2013,4(6):642—646. H,Chadli M,Rabhi A,et a1.Road curvature estimation for vehicle lane departure detection using a robust Takagi— [12] DahmaniSugeno fuzzy observcr[J].Vehicle System Dynamics,2013,5(51):581—599. [13] Yoo H,Yang U,Sohn K.Gradient.enhancing conversion for illumination・robust lane detection[J].IEEE Transactions on Intelligent Transportation Systems,2013,14(3):1083—1094. H,Min J,Kim J.An efifcient lane detection algorithm for lane departure detection[c]//IEEE Intelligent Vehicles [14] JungSymposium(IV)June 23—26,2013,Gold Coast,Australia:976—981. ain D 0,Glavin M,Jnes E.Multiple.camera lane departure warning system for the automotive environment[J].IET Intell [15] CualTransp Syst,2012,3(6):223—234. k S C,Na I S,Han T H,et a1.Lane detection and tracking in PCR gel electrophoresis images[J].Computers and [16] ParElectronics in Agriculture,2012(83):85—91. Yifei,Dahnoun N,Achim A.A novel system for robust lane detection and tracking[J].Signal Processing,2012(92): [17] Wang319—334. ngshu,et a1.Lane detection algorithm for night—time digital image based on distribution [18] You Feng,Zhang Ronghui,Zhong Lifeature of bounda ̄pixels[J].Journal of the Optical Society of Korea,2013,2(17):188—199. zoda R K,Sathyanarayana S,Srikanthan T.Hierarchical additive hough transform for lane detection[J].IEEE Embedded [19] SatSystems Letters,2010,2(2):23—26. adovic D,Konjovic Z,Pap E,et a1.Linear fuzzy space based road lane model and detection[J].Knowledge。Based [20] ObrSystems,2013(38):37—47. iang Ruyi,Reinhard K,Tobi V,et a1.Lane detection and tracking using a new lane model and distance transform[J]・ [21] JMachine Vision and Applications,201 1(22):721—737. (责任编辑李亚青)