Probability Detection of Road Obstacles Combining with Stereo Vision and Superpixels Technology
-
摘要: 针对道路障碍物检测问题,提出了一种基于立体视觉与超像素技术的双目障碍物概率检测算法。该算法首先利用双目相机采集的左右视图进行立体匹配获取视差图,通过三维重建到得3D点云,然后采用最小二乘法拟合地平面方程,并计算图像空间点到地平面的距离,接着使用SLIC算法得到超像素,并计算超像素到地平面的中值距离,最后使用逻辑回归方法和信任函数计算像素为障碍物的概率。实验结果表明,该方法能够可靠地检测道路障碍物,并为障碍物检测的多传感器融合方法提供有价值的概率数据。Abstract: Aiming at the road obstacle detection, the obstacle probability detection algorithm based on the stereo vision and superpixel technology is proposed. Firstly, the disparity map was obtained by using stereo matching on the left and right images captured by using binocular cameras, and the 3D point cloud was reconstructed. Then the least square method was used to estimate the ground plane equation and calculate the distance from the image space point to the ground plane. After that, the superpixels were obtained by using SLIC algorithm, and the median distance of the superpixels to the ground plane were calculated. Finally, the probability of the pixel being an obstacle was calculated by using the logistic regression method and the belief function. The experimental results show that the present method can detect road obstacles reliably and provide the valuable probabilistic data for multi-sensor fusion methods of obstacle detection.
-
Key words:
- stereo vision /
- disparity map /
- 3D point cloud /
- superpixel /
- obstacle detection
-
表 1 不同概率类型准确率和召回率检测结果
概率类型 准确率/% 召回率/% 中值概率 84.5 90.4 可信概率 88.6 86.2 似然概率 82.3 92.8 -
[1] Discant A, Rogozan A, Rusu C, et al. Sensors for obstacle detection-a survey[C]//200730th International Spring Seminar on Electronics Technology, 9-13 May 2007, Cluj-Napoca, Romania. Cluj-Napoca: IEEE, 2007: 100-105 https://www.researchgate.net/publication/4311966_Sensors_for_Obstacle_Detection_-_A_Survey [2] Kaneko N, Yoshida T, Sumi K. Fast obstacle detection for monocular autonomous mobile robots[J]. SICE Journal of Control, Measurement, and System Integration, 2017, 10(5):370-377 doi: 10.9746/jcmsi.10.370 [3] Franke U, Pfeiffer D. The stixel world-a compact medium level representation of the 3D-world[M]//Denzler J, Notni G, Süße H. Pattern Recognition. Berlin: Springer-Verlag, 2009: 51-60 [4] Dan L, Garnett N, Fetaya E. StixelNet: a deep convolutional network for obstacle detection and road segmentation[M]//Xie X H, Jones M W, Tam G K L. British Machine Vision Conference. Swansea: BMVA Press, 2015: 109.1-109.12 [5] Lee T J, Yi D H, Dan D D. A monocular vision sensor-based obstacle detection algorithm for autonomous robots[J]. Sensors, 2016, 16(3):311 doi: 10.3390/s16030311 [6] 林川, 宋伟奇, 覃金飞.基于V-视差的障碍物检测改进方法[J].科学技术与工程, 2014, 14(1):86-90 doi: 10.3969/j.issn.1671-1815.2014.01.018Lin C, Song W Q, Qin J F. Improved method of obstacle detection based on V-disparity[J]. Science Technology and Engineering, 2014, 14(1):86-90 doi: 10.3969/j.issn.1671-1815.2014.01.018 [7] 王军华, 李丁, 刘盛鹏.基于改进RANSAC的消防机器人双目障碍检测[J].计算机工程与应用, 2017, 53(2):236-240 doi: 10.3778/j.issn.1002-8331.1504-0075Wang J H, Li D, Liu S P. Improved RANSAC estimation based fire-fighting robot obstacle detection using binocular vision[J]. Computer engineering and Application, 2017, 53(2):236-240 doi: 10.3778/j.issn.1002-8331.1504-0075 [8] 谭健胜.基于双目立体视觉的三维重建技术研究及实现[D].广州: 华南理工大学, 2017 http://cdmd.cnki.com.cn/Article/CDMD-10700-2008160666.htmTan J S. Research and implementation of 3D reconstruction technology based on binocular vision[D]. Guangzhou: South China University of Technology, 2017 http://cdmd.cnki.com.cn/Article/CDMD-10700-2008160666.htm [9] 罗世民, 李茂西.双目视觉测量中三维坐标的求取方法研究[J].计算机工程与设计, 2006, 27(19):3622-3624 doi: 10.3969/j.issn.1000-7024.2006.19.037Luo S M, Li M X. Research on how to get object's 3D coordinate on two CCD camera measure system[J]. Computer Engineering and Design, 2006, 27(19):3622-3624 doi: 10.3969/j.issn.1000-7024.2006.19.037 [10] 高如新, 朱烜甫.基于ORB算法的双目视觉定位[J].电子测量技术, 2017(4):142-145 doi: 10.3969/j.issn.1002-7300.2017.04.032Gao R X, Zhu X F. Based on the ORB algorithm of binocular vision positioning[J]. Electronic Measurement Technology, 2017(4):142-145 doi: 10.3969/j.issn.1002-7300.2017.04.032 [11] Geiger A, Roser M, Urtasun R. Efficient large-scale stereo matching[C]//10th Asian Conference on Computer Vision. Queenstown, New Zealand: Springer, 2010: 25-38 doi: 10.1007%2F978-3-642-19315-6_3 [12] Mufti F, Mahony R, Heinzmann J. Spatio-temporal RANSAC for robust estimation of ground plane in video range images for automotive applications[C]//11th International IEEE Conference on Intelligent Transportation Systems. Beijing, China: IEEE, 2008: 1142-1148 https://www.researchgate.net/publication/224364932_Spatio-Temporal_RANSAC_for_Robust_Estimation_of_Ground_Plane_in_Video_Range_Images_for_Automotive_Applications [13] Wedel A, Badino H, Rabe C, et al. B-spline modelling of road surfaces with an application to free-space estimation[J]. IEEE Transactions on Intelligent Transportation Systems, 2009, 10(4):572-583 doi: 10.1109/TITS.2009.2027223 [14] Achanta R, Shaji A, Smith K, et al. SLIC superpixels compared to state-of-the-art superpixel methods[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2012, 34(11):2274-2282 doi: 10.1109/TPAMI.2012.120 [15] 张安安, 彭嵩松, 杨威.基于信任函数理论的修正融合目标识别算法[J].电子技术应用, 2015, 41(6):84-87 http://d.old.wanfangdata.com.cn/Periodical/dzjsyy201506026Zhang A A, Peng S S, Yang W. A modified fusion algorithm for target recognition based on the belief function theory[J]. Application of Electronic Technique, 2015, 41(6):84-87 http://d.old.wanfangdata.com.cn/Periodical/dzjsyy201506026 [16] Sentz K, Ferson S. Combination of evidence in dempster-shafer theory[J]. Contemporary Pacific, 2002, 11(2):416-426 http://d.old.wanfangdata.com.cn/Periodical/glgcxb201803014