利用雙眼立體視覺輔助行人偵測的方法

行人辨識一直是物件辨識領域當中一個相當熱門的研究主題，由於交通意外事故一直都是國內十大死因之一，因此，有越來越多的車廠願意將駕駛輔助系統搭載於自家的產品之上，以加強駕駛的安全性，其中包含車道辨識、車輛辨識、以及行人辨識都是相當重要的研究領域，使得駕駛人可以更安心上路。駕駛輔助系統之人物辨識對於準確率以及即時性的要求相當高，因此，本論文提出以兩個CCD 攝影機的雙眼立體視覺產生視差圖，根據視差圖之深度資訊劃分前景物與背景物，再利用障礙物於三維空間之物理特性切割出我們感興趣的區域，最後擷取出行人之梯度方向直方圖特徵，並經由支援向量機分類驗證行人影像，成功分類出行人影像後，我們可以更進一步的利用雙眼立體視覺當中的三角幾何關係計算出行人距離。經由實驗驗證，我們的方法可以達到更快速且正確的辨識結果，以及在距離估測上的準確性也有相當不錯的效果。

關鍵字

雙眼視覺；行人辨識；支援向量機；梯度方向直方圖

並列摘要

Pedestrian detection is always a popular research topic in the objective detection area due to transportation accident is one of the main causes of death in Taiwan. Hence, more vehicle companies are willing to accompany driving assistance system in their own products in order to increase the safety of driving. In addition, lane detection, vehicle detection, and pedestrian detection are significant research fields to allow safer driving. Accuracy rate and real time processing are highly required in the pedestrian detection of driving assistance system, therefore, our essay suggests to use two CCD camera to build stereo. Based on the depth information of stereo, we can divide foreground and background.Following, the obstacle of physical property of three-dimensional space are used to define and segment the region that we are interesting to capture the histogram of oriented gradient feature of pedestrian. Furthermore, through support vector machine classifier to verify the pedestrian image. Finally, we calculate the pedestrian distance by using the stereo of triangular geometric relations. According to our experiments, this method is able to achieve a faster and more correct classify result. As well as a remarkable result in accuracy of distance estimation.

並列關鍵字

stereo vision ； pedestrian detection ； support vector machine(SVM) ； histogram of oriented gradients(HOG)

參考文獻

[1] S. H. Lai, C. W. Fu, S. Chang, “A generalized depth estimation algorithm with a single image”, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 14, No. 4, pp. 405-411, 1992

[2] C. R. Wren, A. Azarbayejani, T. Darrell, A. P. Pentland, “Pfinder: real-time tracking of the human body”, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 19, No. 7, pp. 780-785, 1997.

[3] R. Cutler, L. S. Davis, “Robust real-time periodic motion detection, analysis, and applications”, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 22, No. 8, pp. 781-796, 2000.

[5] Yajun Fang, K. Yamada, Y. Ninomiya, B. K. P. Horn, I. Masaki, “A shape-independent method for pedestrian detection with far-infrared images”, IEEE Transactions on Vehicular

[6] R. Y. Tsai, “A versatile camera calibration technique for high-accuracy 3D machine vision metrology using off-the-shelf TV cameras and lenses”, IEEE Journal of Robotics and

國際替代計量

利用雙眼立體視覺輔助行人偵測的方法

全文下載

主題瀏覽