Construction and study of an improved PP-Human multi-objective tracking method

Abstract

Multiple objective tracking is an important research direction in the field of computer vision. Efficient multitarget tracking method is of great significance for applications such as video surveillance, unmanned driving and intelligent security. In complex environments, due to interference such as light and noise, low-light image enhancement instability, target correlation accuracy and robustness are faced. In this paper, PP-Human is used for multi-objective tracking. By training and optimizing PP-Human model, combining high-precision detector, strengthening pedestrian reidentification (ReID) technology and optimizing data association strategy, the PP-Human model improves multi-object detection capability and realizes efficient and accurate multi-objective tracking. Evaluation the effect of the present multi-target tracking method using the enhanced Market 150 dataset improves the accuracy of multi-target tracking by 2.5% to 95.0mAP, verifying the effectiveness and robustness of the present method. Through experimental verification, the improved PP-Human performed well in multitarget tracking tasks, providing a solid foundation for pedestrian analysis, behavior recognition, and flow statistics in practical applications. Note to Practitioners (NtP)—In the realm of computer vision, multi-objective tracking poses significant challenges, especially in complex environments where lighting conditions and noise interference can affect performance. Our research addresses these challenges by leveraging the PP-Human model for more effective and precise tracking. Practitioners in fields like video surveillance, unmanned driving, and intelligent security stand to benefit greatly from our work. The enhanced PP-Human model not only boosts multi-object detection capabilities but also ensures more accurate and efficient tracking. This is achieved through a combination of a high-precision detector, strengthened pedestrian reidentification (ReID) technology, and an optimized data association strategy. Our method, evaluated using the enhanced Market 150 dataset, demonstrates a 2.5% improvement in multi-target tracking accuracy, achieving a remarkable 95.0mAP. This validation underscores the method’s effectiveness and robustness, making it a viable solution for real-world applications. It is worth noting that the improved PP-Human model excels in multi-target tracking tasks, laying a solid foundation for further pedestrian analysis, behavior recognition, and flow statistics in practical settings.