快速消除车辆阴影的多阈值图像分割法

 视频图像分割时的运动阴影由于与被测对象的相似性而常被误判为被测对象，传统的阴影检测方法一般难以满足实时智能交通系统对处理速度的要求。为此，提出用多阈值法分割图像，将灰度化的当前图像与背景差分，再用正、负两个阈值对其二值化，在分割出深色和浅色被测对象的同时消除阴影。实验表明，将这种方法应用于运动车辆的检测，具有速度快和阴影消除效果好等特点，可应用于实时的运动目标检测和跟踪等领域。

Quick elimination of vehicle shadow by multi-threshold image segmentation

Moving shadow is usually misinterpreted as part of moving object because of its similarity with the object in video target segmentation, and the processing speed of traditional shadow elimination method can not meet the requirement of real-time intelligent transportation system. A new shadow elimination method is proposed. In this method, every color image of a video is transformed into gray images; and difference image is obtained by calculating the difference between the gray image and its background; then the difference images are transformed into binary images with a positive threshold and a negative threshold. By selecting these two thresholds appropriately for both light-colored and deep-colored objects, only target object remains in the binary images. In the experiments of real-time vehicle detection, this method demonstrates a good performance of quick shadow elimination compared to traditional method. It can be applied to moving object detection, tracking and other real-time applications.