As sudden illumination change occurs i n video sequences current,moving target detection algorithms provide low accurac y and higher false results,so an algorithm combined with illumination compensation is proposed in this paper.At first,stable background areas are extracted by app lying chromatic invariants, and illumination change degree is determined by combining with the history of moving objects.Then, the image sequences are compensated by the diagonal model,and the model parame ters are respectively calculated through interval estimation and codebook cluster for the global and local illumination changes.Moving objects detection is achieved by background subtraction finally. Experimental results show that the proposed algorithm is applicable to moving object detection in dyn amic scenarios and provides higher detection accuracy and stability.