As the existing background subtraction algorithm is insufficient for the real-time expressway monitoring with frequent light change，an adaptive threshold determination algorithm based on the gray level is proposed． In this algorithm，the pixels of the object in the difference image are effectively classified by means of the statistical method，and the classification standard is defined as the adaptive threshold． Moreover，the gradient distribution of the difference image is used to judge the area of motion objects． Experimental results show that the proposed adap-tive threshold algorithm is effective in recognizing objects in different environments with high accuracy and stability．