SLR Video Weak Target Detection Based on Spatial-Temporal Variance Fusion

To address the critical detection challenges in Satellite Laser Ranging (SLR) observations-including random noise, fixed-pattern interference (e.g., hot pixels), dynamic trajectory confusion, strong signal masking, and low grayscale contrast-this paper proposes a weak target detection algorithm for SLR videos based on the fusion of spatial and temporal variance.First, a comprehensive preprocessing module is constructed, incorporating bad pixel mask restoration and column-wise statistical differencing to concurrently eliminate intrinsic sensor hot pixels and vertical stripe noise; meanwhile, a non-linear square mapping is employed to enhance the grayscale contrast of weak targets. Second, a dual-channel detection mechanism is designed by leveraging the complementary characteristics of spatio-temporal variance: the "spatial variance channel" utilizes temporal mean and spatial variance to jointly accumulate weak signals, addressing the "invisibility" issue caused by single-frame noise; the "temporal variance channel" utilizes the temporal fluctuation features of pixel intensity to physically distinguish moving targets from static high-brightness interference, thereby resolving the "indistinguishability" problem. Finally, the strong signal masking effect is effectively overcome through the fusion of spatio-temporal variance features and a dynamic saturation mechanism based on the noise floor.Extensive tests on 92 real observation videos involving 12 types of satellites demonstrate that the proposed algorithm achieves a high detection rate of 95.6% while suppressing the average number of false alarms to 0.03. The experimental results indicate that the algorithm can effectively mitigate the effects of strong interference and low contrast, achieving highprecision localization and extremely low false-alarm detection for weak SLR targets.