Real-Time Infrared Target Tracking on Embedded Platform Based on Improved Staple Tracker

Infrared target tracking aims to continuously obtain the position and size of specific infrared target in image sequences or videos. Existing tracking methods based on correlation filter are limited to the spatial texture of infrared target, which cannot tackle completely insufficiency of feature information in tracking process. To solve the problem, we propose a real-time infrared target tracking method in this paper, which combines the probability map of edge features and Staple tracking algorithm. Specifically, the proposed method enhances the edge details between infrared target and background and learns the deformation of infrared target. To cope with occlusion during infrared target tracking, we incorporate a template update module with average peak correlation energy to detect the occlusion of infrared target and selectively update the tracking template model. The experimental results show that the improved method achieves satisfactory improvement compared with the original Staple algorithm in the tracking precision and success rate. Furthermore, we also achieve real-time running on the embedded platform.