基于快速小波变换和卡曼滤波的车辆检测与轨迹跟踪

Vehicle Detection and Tracking Based on Fast Wavelet Transform and Kalman Filter

主要研究一种新的车辆检测方法。在图像处理系统中,首先通过快速小波变换提取图像的纹理,同时利用灰度共生矩阵对提取出的纹理进行度量与分析。在此基础上根据图像中各部分的纹理差异检测车辆的存在,并成功剔除阴影的干扰。其次,提出用于图像处理的卡曼滤波的状态转移矩阵和观测矩阵,并利用其对车辆的状态进行跟踪,从而达到在图像序列中跟踪车辆运行轨迹的目的。实际道路环境下的实验充分说明所提出的方法的有效性。

A new approach for vehicle detection is proposed in this work. First, fast wavelet transform （FWT） designed for discrete signal is proposed to extract image texture, while grey level co-occurrence matrix （GLCM） is employed to measure and analyze the extracted texture. Then, vehicles can be extracted since the vehicle sections and the shadow sections have different textures in the combined foreground image. Moreover, we put forward in this work the state and observation matrixes of Kalman filter which has been employed to track vehicles under complicated traffic scenes. Experimental results in real traffic scenes reveal that the proposed techniques are effective and efficient for vehicle detection and tracking.