Accurate and efficient moving object detection with ViBe and visual foreground contour extractor

Moving object detection is to perform a series of analysis and operation on video or image sequence, calculate the position of moving object in the image, and extract the moving object from the complex background, so as to obtain an image containing only moving objects. The ViBe algorithm is a commonly used method, but when using ViBe algorithm to recognize and detect moving objects in video or image sequences, there is a “ghost” phenomenon and it is suffer lots of environmental noises. Therefore, a Visual Foreground Contour Extraction algorithm is designed to improve ViBe. When background modeling, the Visual Foreground Contour Extraction algorithm uses the mode background method to establish the initial background and background sequence; In foreground detection, background difference method and Sobel operator are used to calculate the moving object area, and adaptive multilevel threshold is used to denoise; Finally, the Visual Foreground Contour Extraction algorithm is intersected with the ViBe algorithm, and the complete moving target is obtained by mathematical morphology processing. At the same time, the parallel scheme of CPU-GPU is designed, which uses CPU parallel to calculate image background and GPU to accelerate foreground detection. The algorithm is tested on the CDNet2014 datasets. The experimental results show that the precision, recall rate and F1 score of the algorithm are 32.14%, 9.64% and 20.76% higher than ViBe. The missed detection rate and false detection rate are low and the accuracy is high; In terms of efficiency, the average detection efficiency of the algorithm is 64.70% higher than ViBe, and has good real-time performance.