Abstract: In view of the current number of lift platform overload detection is given priority to with artificial, man-made interference, disadvantages are more obvious. In this paper, a number of lift platform based on computer vision image statistical method. Using discrete wavelet transform algorithm to deal with the noise to collect images of lift platform, using the gaussian curve fitting method to image feature extraction, head using the head and the background grey value change is continuity of contour area. Workers contour with the contour of a background texture feature differences. Workers contour edge pixel gray value exists gradient change these three characteristics, to division head, according to the characteristics of personnel area marked segmentation result and counting. The experimental results show that the use of the method can accurately on the number of people on the elevator platform statistics, error less than 5%, can meet the demand of practical application.