基于对称性的快速车辆检测方法

Symmetry-based vehicle detection method for rapid

提出了一种基于车辆边缘对称性和水平阴影的快速车辆检测方法。通过利用被检测图像减去道路背景图像的方法对被检测图像进行预处理,以最大限度去除背景图像中车道线等大量干扰,突出车辆图像,对图像分别做Sobel垂直和水平边缘检测。把垂直边缘图像分割放缩成远景、中景、近景3幅大小相同的图像,利用车辆对称性对分割放缩后3幅图像对称性进行检测,确定图像中所有车辆在图像中的水平位置。同时在水平边缘检测的图像中,利用水平阴影和汽车底部水平边缘确定汽车的垂直位置。实验结果表明,该方法简单、有效,能快速、精准地检测出汽车图像中所有的不同位置和不同大小的车辆。

A fast vehicle detection method based on symmetry of vehicle edge and the level of the shadow is proposed. Firstly, the detected image through the use the method which makes the detected image minus the background image, and remove large number of disturbances in the lane maximum in the background image to highlight the vehicle images Secondly, the edge detec- ting of the image with Sobel, vertical and horizontal edges detection algorithms separately is finished. And then zoom and seg- ment image to three of the same size image like vision, medium shot, close range image. Finally, the symmetry of the vehicle is used to detect the three separate images and make sure all vehicles at the image＇s horizontal position in the image, and the level of the shadow and the level edge of the bottom of the car are utilized to determine the vertical position of the car. Experimental results show that this method is simple, rapid and can detect out all the different locations of vehicles and different sizes of the vehicle in the image accurately.