Abstract: An algorithm based on adaptive background and improved dynamical threshold is proposed to solve the problem of the limitation of current detection methods of moving vehicle in complex scenes. Obtaining the difference image by the substraction between present frame and background, the precision detection of moving objects is implemented through binaryzing the three different color channels of the difference image by selecting adaptive threshold values. According to the detected objects, an updating strategy by mean values is introduced to achieve real-time update of background. Experimental results of outdoor image sequences demonstrated that the proposed algorithm can effectively detect moving objects in complex traffic scenes, and what's more its computation cost is reduced with higher robustness and real-time performance provided, thus meeting the requirements of real-time detection of moving vehicles in intelligent transportation surveillance system.