Moving target detection based on PERT background model

Moving object detection is one of the important researches in computer vision. The traditional background subtraction approaches need to get an ideal background image which does not include any moving target. However, it is difficult to obtain an ideal background image in reality. The main idea of updating background model is using an update rate to reflect the impact of outside scene changes on the background image, which has less impact of environment changes on the background image. However, the initial background image affects the updating of the background greatly. To solve these problems, the algorithm based on PERT background model is proposed in this paper, where the pixel gray value is assumed to follow beta distribution and the background model is created and updated by the three-time estimation method. To minimize the effects from illumination changes, it is necessary to compensate the detection results with dynamic factor. Experimental results show the algorithm is more effective and reliable.