Adaptive Convolutional Feature Fusion Target Tracking Algorithm based on Spatio-Temporal Regularization

This paper proposes an adaptive fusion convolutional feature tracking algorithm based on the spatio-temporal regularized tracking algorithm to address the issue of instability in tracking models under complex scenarios such as lighting and scale variations. The algorithm in this paper replaces traditional handcrafted features with convolutional features from different layers of a convolutional neural network, assigning different weights to these features. It utilizes stable spatio-temporal regularized correlation filters for target estimation and continuously updates the weights to enhance the stability of the model in different scenes. The proposed algorithm is tested on the OTB-2015 dataset. Experimental results demonstrate that the algorithm in this paper achieves good accuracy and stability, with favorable performance in terms of precision and success rate across all video sequences.