Improved ViBe algorithm based on multi-frame combined with adaptive threshold

Abstract

For moving human target detection under video surveillance, the traditional vibe algorithm is often disturbed by environmental changes, resulting in “ghosts”, incomplete detection results and “holes” in the human body. An improved vibe algorithm based on multi frame combined with adaptive threshold is proposed. Firstly, the sample set of vibe algorithm is expanded to 24 fields to reduce the possibility of pixel misclassification; Secondly, the historical pixel queue is introduced, and the initialization background model without ghost is obtained according to the change of foreground and background pixels in time domain; Finally, the distance determination threshold is dynamically adjusted by using the gray characteristic convergence and divergence, and the adaptive update factor of the model is calculated by introducing two fixed parameters to optimize the update rate of the background model. The experimental results show that the algorithm is not only suitable for general scenes, but also for dynamic complex scenes, it can accurately detect the foreground of moving human targets, eliminate “ghosts”, obtain complete moving targets, avoid the appearance of “holes”, improve the accuracy of detecting foreground human targets in dynamic complex environments, and enhance the robustness of the algorithm.