基于Mean Shift方法的视频车辆检测与分割

提出了一种基于Mean Shift方法的视频车辆检测和分割方法.首先将交通场景图像与路面区域所对应的二值掩模图像进行“与”运算以排除无关背景干扰,并对所得的结果图像用Mean Shift聚类方法进行分割以得到原始的分割图像.然后根据区域的面积、分布以及颜色的均匀性和相似性等特征有效过滤出路面区域.进一步基于颜色的不相似度量,将路面区域置“黑”,所有其他区域置“白”,对路面区域图像进行二值化.最后通过特定的后处理过程可把路面区域中所存在的动、静态车辆检测出来.

Video vehicle detection and segmentation based on Mean Shift method

A video vehicle detecting and segmenting approach is introduced based on the Mean Shift method.First,it eliminates the disturbance from unrelated background by AND operation between a traffic image and the binary mask of road area,and then it segments the result image by Mean Shift clustering method to obtain the original segmented image.The road area can then be filtered out according to the features such as area,distribution of the segmented regions,uniformity and similarity of region colors,and so on.Further,based on dissimilarity measurement of the color,the region of road area is set to be "white" and all the other regions "black" to obtain the binary image.Finally,both static and moving vehicles can be detected by certain post-processing steps.Validity of the method is proved by the experimental result given at the end.