UAV Cluster Detection Algorithm Based on Weighted Tri-Layer Window Local Contrast

Infrared images of unmanned aerial vehicle (UAV) cluster targets are characterized by the low signal-to-noise ratio (SCR), closely spaced, and limited detail characteristics, which exacerbate the detection complexity of contemporary infrared (IR) small target detection algorithms. This paper proposes a weighted tri-layer sliding window local contrast (NTLLCM) algorithm for UAV cluster targets detection. Firstly, the IR imaging characteristics of UAV cluster targets are modeled and evaluated. Secondly, after evaluating the limitations of LCM, the local contrast of the filtered images is calculated by designing a tri-layer sliding window structure with core, middle, and outermost layers. The local contrast is then weighted utilizing a new regional intensity level (NRIL) strategy to suppress the background further, and the target is detected by adaptive threshold segmentation. Finally, a considerable number of experimental results illustrate the superiority of this algorithm in detecting UAV cluster targets under sophisticated backgrounds, especially the proposed algorithm detects multiscale targets utilizing single-scale windows instead of using multiscale windows, which significantly decreases the computationally costs. Furthermore, the algorithm can reduce the overlapping effect of UAV cluster target imaging and correctly detect closely spaced targets.