Infrared Small Moving Target Detection Based on Spatial-Temporal-Frequency Analysis

The Infrared Searching and Tracking System (IRSTS) plays an important role in both military and citizen life, contributing to crucial areas such as security surveillance and transportation, and so on. As the core step in the system, methods for infrared moving target detection and tracking attract much attention in recent years, especially for several specific small targets like unmanned aerial vehicles (UAVs). In this work, a spatial-temporal-frequency feature fusion based approach is proposed. Firstly, the infrared sequential images are decomposed via 3D wavelet transform. Only the high frequency information in spatial and temporal domains is reconstructed to form the reconstructed sequence. A background suppression strategy from temporal to spatial domain is secondly applied in the reconstruction images and the target is acquired by adjust segmentation. Experimental results demonstrate that the proposed method achieves competitive performance both in qualitative and quantitative comparisons.