Long-term infrared object tracking algorithm based on dynamic region focusing for anti-UAV

The misuse of unmanned aerial vehicles (UAV) is accelerating the development of anti-UAV technologies. Infrared detector-based tracking methods have gained special attention in the anti-UAV field, which, however, still face the problem of tracking failures caused by background interference. To enhance the precision and stability of infrared anti-UAV tracking in complex environments, this paper proposed a long-term infrared object tracking algorithm based on dynamic region focusing. Firstly, the Siamese backbone network based on feature pyramid was constructed to improve the feature extraction capability of the model for infrared UAV by the fusion of cross-scale features. Secondly, a dynamic region proposal network based on spatio-temporal joint constraints was proposed. Under the constraints of template appearance features and target motion information, the location probability distribution of the object was predicted over the entire image, and then the prior anchor box was guided to focus on the candidate regions, realizing a dynamic search region selection mechanism. The anti-background interference capability of local search and the recapture ability of global search were subtly integrated by focusing on the search area, which effectively mitigated the negative sample interference caused by global search and further enhanced the discriminability of target features. Experiments on the Anti-UAV dataset show that the proposed algorithm achieves precision of 0.895, a success rate of 0.649, and average accuracy of 0.656 with a tracking speed of 18.5 FPS. Compared with other advanced tracking algorithms, the proposed algorithm exhibits superior performance and demonstrates its effectiveness in handling complex tracking scenarios such as fast motion, thermal crossover, and similar distractors.