Object Tracking Algorithm Based on Margin Feature Fusion

In the target tracking algorithm, extracting features from images is a vital link. By extracting the features of the tracking target and the surrounding area samples, the higher the similarity is, the greater the correlation of the image is, so as to distinguish the target from the background, and then determine the location of the tracked target. How to extract the effective feature information from the image directly affects the detection effect of the algorithm. Most target tracking algorithms adopt HOG feature and grayscale feature to extract target shape feature. When the target has local occlusion, illumination change, the target deformation, these characteristics of the target expression is not substantial, easy to appear tracker drift, thus reducing the accuracy of target tracking, and even lead to the occurrence of tracking failure. In order to solve this problem, this paper proposes a target tracking algorithm that integrates edge features, fuses edge features with richer textures and HOG features of the performance space, so as to improve the performance of the extracted features. In the case of target deformation, local occlusion and illumination change, the image features can also be extracted effectively, the model can be updated, and the target can be tracked. Experimental results show that the target tracking algorithm based on edge features is more stable than some traditional correlation filtering algorithms.