Sports Video Target Tracking Algorithm Based on Optimized Particle Swarm Algorithm

Abstract

With the rapid development of computer technology and the continuous improvement of sensor performance, the role of sports video target tracking technology in practical applications such as intelligent sports competitions and international sports stadium security monitoring systems has become increasingly prominent. It has become the current research hotspot and difficulty in the field of computer sports vision and artificial intelligence. In order to achieve long-term, stable, accurate and efficient tracking of targets in complex environments, it is of great significance to establish an effective adaptive model and research on sports video target tracking algorithms. The purpose of this paper is to study the sports video target tracking algorithm based on the optimized particle swarm algorithm. This paper proposes a particle filter video target tracking algorithm based on optimized particles. Based on the particle filter framework, this algorithm optimizes the particles before resampling by using the particle swarm optimization algorithm, keeping particles with larger weights still, and allowing particles with smaller weights to continuously approach particles with larger weights, so that most of the particles are in the high-likelihood area, thereby reducing the particle elimination rate in the re-sampling process and improving the problem of particle shortage. This paper uses the least squares-based trajectory prediction intervening particle filter algorithm at the level of the algorithm process framework, which makes the improved tracking algorithm more robust in dealing with target occlusion, especially in the case of long-term occlusion and background mutation. Experimental research shows that at the 27th frame, because the target is temporarily occluded by the foreground person, when the target reappears from behind the occluded foreground, the target cannot be tracked correctly, but the foreground occlusion is mistakenly regarded as the target for tracking iterations leading to the tracking process Failure, because the target keeps moving forward while the foreground target remains relatively stationary, so the tracking error percentage gradually increases.