In the RT-BSLSAC model, the image information of adjacent frames is used to detect the moving object in the current frame. Through reasonable design of model and reasonable selection of numerical algorithm, higher segmentation accuracy and better real-time performance are obtained. However, due to the lack of direct detection of image edges, the RT-BSLSAC model may still not be accurate enough to locate the human body contour. In addition, the numerical calculation accuracy is easily affected by the value of the initial LSF, which is caused by the non-convex model energy functional. . This section discusses this problem in order to improve the segmentation accuracy of the model and the robustness to the initial conditions. The improved model is called the accuracy-enhanced background subtraction level set active contour (the background sub extraction based LsAC with enhanced accuracy, EA- BSLSAC) model.