A New Approach for Moving Shadow Detection

Abstract

In the process of traffic video segmentation, the moving shadows have always been wrongly detected as foreground, which will in turn affect the latter processes of detection, analysis and tracing. We propose a method of moving shadow detection based on local texture features integrated with linear characteristics. Firstly, an adaptive background model was built on the basis of RGB color information, and speedy moving target detection was based on the background subtraction. And then, do shadow detection through Hamming distance by using the texture information which described by improved LBP local texture operator. Finally, the judgement of linearity occur via the previous texture information for the further shadow detection. According to the experiment, this method succeeds in detecting shadows robustly and achieves real-time performance.