混合交通环境中的阴影检测算法

在城市交通流量视频检测系统中,目标阴影总是干扰对目标的正确检测和识别。在混合交通环境下,传统的阴影检测算法总是避免不了进行边缘检测、模板匹配等运算,不仅处理速度慢,而且对行人阴影的检测效果不好。本文提出一种基于颜色信息的阴影检测算法,该算法首先在图像中检测出运动区域,然后在运动区域内计算目标R、G、B颜色分量的灰度距离和色彩距离;最后根据这两个距离量检测出区域中的阴影。实验表明,该算法能够正确检测出车辆和行人的阴影,还能在雨天对目标的路面倒影进行检测,而且计算速度较快。

The Shadows Detection Algorithm in Multi-Traffic Scenes

In the video detection system of urban traffic flow, the existence of moving cast shadows always leads to an inaccurate object detection and recognition. In multi-traffic scenes, the conventional algorithms always use edge detection and model match to detect shadows. Using these conventional algorithms, the process speed is slow and results are unsatisfied. This paper presents a novel algorithm for detecting shadows using color information. Firstly, the algorithm detects the motion regions of the image. Then it computes RGB gray-value distance and coloring distance of the moving regions. Finally, the shadows are separated from the regions using the distance value. As shown in experiments, in multi-traffic scenes, the algorithm can detect shadows of various vehicles and pedestrians, even the inverted reflection of objects in the wet road surface under the rainy conditions, and the process speed is high.