EASATrack: Infrared Object Tracking Based on Edge Activation and Scale Adaptive Cross-Correlation.

The blurring of object edges in infrared images and scale variations among consecutive frames largely affect the precision of existing tracking model. To this end, a model called EASATrack based on edge activation and scale adaptive cross-correlation is designed. First, in this model, an edge activation module is employed to amplify the edge information of the object within infrared images. Then, a scale adaptive cross-correlation module is designed to perceive object information across varying scales, thereby accommodating changes in object scale during similarity computations. Finally, to validate the proposed model, plenty experiments are conducted on publicly available datasets PTB-TIP and LSOTB-TIR. The results demonstrate that the proposed model achieves leading performance in terms of precision and success rate on both datasets.