论文部分内容阅读
提出了一种基于单双目视觉融合的车辆检测与基于Kalman滤波的车辆跟踪算法,设计了一种基于二维深度置信网络的车辆检测器。在道路图像中利用单目视觉生成车辆可能存在的区域,构成双目视觉处理的车辆候选集合。在车辆可能存在的区域内利用双目视觉进行误检去除,并获得车辆的位置信息。在二维图像坐标系和三维世界坐标系内,利用Kalman滤波器对检测到的车辆进行跟踪。试验结果表明:算法的检测率为99.0%,误检率为1.3×10-4%,检测时间为57ms,检测率高,误检率低,检测时间短;与单双目视觉弱融合算法、单目视觉算法和双目视觉算法相比,本文车辆检测与跟踪算法兼具双目视觉算法检测率高和单目视觉算法检测时间短的优点。
A vehicle detection algorithm based on single-binocular visual fusion and a vehicle tracking algorithm based on Kalman filtering are proposed. A vehicle detector based on two-dimensional depth-based confidence network is designed. In the road image, monocular vision may be used to generate a region where vehicles may exist to form a binocular vision candidate vehicle candidate set. Use binocular vision for false detection in areas where vehicles may be present and obtain vehicle location information. In the two-dimensional image coordinate system and the three-dimensional world coordinate system, the Kalman filter is used to track the detected vehicles. The experimental results show that the detection rate of the algorithm is 99.0%, the false detection rate is 1.3 × 10-4%, the detection time is 57ms, the detection rate is high, the false detection rate is low and the detection time is short. With weak binocular vision fusion algorithm, Compared with binocular vision algorithm, the vehicle detection and tracking algorithm in this paper combines the advantages of high detection rate of binocular vision algorithm and short detection time of monocular vision algorithm.