Design and Optimization of Visual Tracking Algorithm Based on Deep Learning

Visual tracking is a challenge in computer vision. Visual tracking has various practical application scenarios, and is used in intelligent video surveillance system, analysis and research of driver’s abnormal behavior, intelligent robot research and other fields. Visual tracking algorithm combines video processing, automatic system control, informatics and so on, so as to automatically identify the target from the video image and extract the tracked target position information. In this paper, the visual tracking algorithm is designed and optimized based on deep learning. The experimental data show that the algorithm used in this paper is more advantageous, and the experimental results are evaluated by accuracy and success rate. The figure shows the percentage of the tracked target in the tracking frame. The error between the center position of the tracked target and the real target is measured by different thresholds. Similarly, the success rate indicates the percentage of the successfully tracked target frame, and the proposed algorithm is superior to the traditional tracking algorithm, so it is feasible to optimize the visual tracking algorithm. The target detection algorithm based on deep learning can achieve better target detection effect, and the amount of calculation is small. Continuous real-time tracking can effectively solve the problem of target blocking and improve the accuracy of visual tracking.