UAV Target Tracking Algorithm for Ground Moving Targets

With the goal of effectively estimating the target's location, an algorithm for adaptive target tracking that based on target's response is put forth. First, the extracted features are compressed by an encoder and a soft quantizer. Secondly, the target response regular term is added to the target training model, and the optimal target response function is updated in the process of solving, so that the tracking algorithm is able to precisely track the target after being disturbed, so as to improve the representation ability of the target. Finally, reliability coefficient is introduced to evaluate the response graph of each feature channel to express the ability of the model to express the target. According to experimental data, this method can achieve 94.5 % tracking accuracy.