Moving Vehicle Shadow Detection Approach Based on Region Growing and Gradient Statistics
MA Xin ZHANG Wenjia · Chuangan jishu xuebao · 2010
To prevent the moving shadow from being misunderstood as a part of moving objects in video target segmentation,a novel approach is proposed for the shadow detection,in which the model and the characteristics of the shadow are merged.Using the clustering algorithm to extract the background,then the foreground with the shadow can be obtained.Calculating the gradient difference of the foreground and background and carrying out the projection of the gradient,the initial detection of the shadow region can be obtained according to the sequence and the difference sequence after the projection.Based on the preliminary detection results,the region of the shadow can be searched by using the region growing algorithm.Experimental results show that the algorithm has a good performance,and requires fewer time.It can be used in the real-time detection of the vehicle shadow.