Integrated moving cast shadows detection method for surveillance videos

Moving cast shadow detection plays an exceedingly important role in content-based analysis in surveillance videos. A novel integrated moving cast shadow detection method is proposed for accurate moving objects detection. Based on the analysis for shadow model, we first exploit color information sufficiently in hue, saturation, and value color space and texture similarity in texture space to separate moving cast shadows from the foreground, respectively. Then we obtain the rough shadow detection result by means of synthesizing the above two results efficiently. Finally, post processing is put forward to modify misclassified pixels for acquiring the refined shadow detection result. Extensive experiments and comparisons of various scenes prove that the proposed method can detect cast shadows automatically and effectively and outperforms some well-known methods.