前車偵測以及追蹤的自動駕駛輔助系統是一個十分困難的課題.這個駕駛輔助系統可以偵測前方多台車輛,並且同時追蹤多台車輛.除了這個優點,我們的系統還可以在多種天氣下面來執行,例如:晴天,陰天,雨天.

在我們的系統中,一開始是先使用粒子濾波器(PARTICLE FILTER)偵測出前方車輛有可能出現的位置,利用機率分布的方式,來找出機率最大的所在地,然後再次的判斷出,前方是否有車輛,若車輛出現在前方,則可以將此車輛資訊交給車輛追蹤系統.

當車輛追蹤系統開始追蹤車輛的同時,會利用向量支持機(SVM)來搜尋出每一張影像中最佳的車輛位置,並且將這個最佳位置作為下一張圖片起始搜尋點,藉由此來找出下ㄧ張影像最佳的位置,並且會統計車輛移動軌跡跟大小變化,藉此來提醒駕駛員要與前車保持安全距離.

而我們利用這個搜尋方法將其擴展到多台車輛偵測以及追蹤系統上,首先將輸入的影像區分成三個影像區間,並且分別在三個影像區間上分別做出偵測以及追蹤.左右兩邊的影像區間是用來偵測是否有超車情況,而中間的影像區間是利用來監視是否有前車行駛在本車正前方.利用此系統執行方式.我們就可以擴展到追蹤三台車輛.

如果追蹤系統出錯的時候,則系統會回覆到偵測狀態,藉由此方法來保持系統穩定度.

Preceding vehicle detection and tracking is an important task for vision-based automatic guidance system. In this thesis, we propose a method to solve the problem which includes multi vehicle detection and multi vehicle tracking. In multi vehicle detection, we sample the image using the particle filtering theory. After sampling we are using the SVM to find the vehicle image. In multi vehicle tracking hypothesis, we find the local maximum SVM score to identify the positions of the vehicle. In the experiments, we demonstrate that our system can identify the existing preceding car as well as the passing car. Once the vehicle is detected, it will be tracked until the size of it disappear or occluded by other vehicles.

Reference:
[1] A. Bensrhair, M. Bertozzi, A. Broggi, P. Miche, S. Mousset, and G. Toulminet. “A cooperative approach to vision-based vehicle detection,” IEEE Conf. of Intelligent Transportation Systmes, Oakland USA, Aug. 25-29, 2001.
[2] J. Miura , M. Itoh, and Y. Shirai, "Towards Vision-Based Intelligent Navigator: Its Concept and Prototype", IEEE Trans. on Intelligent Transportation Systems, Vol. 3, No. 2, pp. 136-146, 2002.
[3] N. Shimomura, K. Fujimoto, T. Oki, and H. Muro, “An Algorithm for Distinguish the Types of Objects on the Road Using Laser Radar and Vision,” IEEE Trans. on ITS, Vol.3, No.3, Sept. 2002.
[4] N. Srinivasa, ”Vision-based vehicle detection and tracking method for forward collision warning in automobiles,” IEEE Intelligent Vehicle Symposium, 2002. Volume 2, 17-21 June 2002 PP:626 - 631
[5] N. Srinivasa, Y. Chen, and C. Daniel1, “A fusion system for real-time forward collision warning in automobiles,” IEEE Conf. on Intelligent Transportation Systems, pp. 457 – 462, vol. 1, 12-15 Oct. 2003
[6] T. Kato, Y. Ninomiya, and I. Masaki, ”Preceding vehicle recognition based on learning from sample images,” IEEE Transactions on Intelligent Transportation Systems, Vol. 3, Issue 4, Dec. 2002 , pp:252 – 260 .
[7] X. Li; X. C. Yao; Murphey, Yi.L.; Karlsen, R.; Gerhart, G., “A real-time vehicle detection and tracking system in outdoor traffic scenes,” ICPR 2004. Volume 2, 23-26 Aug. 2004 Page(s):761 - 764 Vol.2
[8] H. Y. Chang, C. M. Fu and C. L. Huang, “Real-time vision-based preceding vehicle tracking and recognition” IEEE Intelligent Vehicle Symposium, 2005. pp. 514-519, Las Vegas, June 2005.
[9] Z. Sun, G. Bebis, and R. Miller, “On-road vehicle detection using Gabor filters and support vector machines,” International Conference on Digital Signal Processing, Greece , July, 2002.
[10] Z. Sun, R.Miller, G. Bebis, D. DiMeo, “A real-time precrash vehicle detection system,” IEEE Workshop on Applications of Computer Vision, 2002.
[11] S. Avidan, “Support vector tracking,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Volume 26, Issue 8, Aug. 2004 Page(s):1064 – 1072.
[11] V. Vapnik, The Nature of Statistical Learning Theory. New York: Springer, 1995.