Study On High Reliability Detection Method Of Infrared Dim Target In Complex Space Environment

The high reliability detection of infrared dim targets in complex space environment is of great significance to the application of infrared staring imaging system in the field of space security.In this paper,based on the imaging model of infrared dim target,the background prediction of infrared image,trajectory prediction of target motion and target detection are studied in detail.In this paper,the geometric model,the detection model and the noise source of infrared dim target are detailed analyzed.According to the moving velocity of the target on the focal plane,the target is divided into stationary target,sub-pixel moving target and pixel moving target.Using multi-frame accumulation method to increase the target signal to noise ratio,improve the target detection performance.Aiming at the problem that the traditional bilateral filter background prediction method cannot change the background prediction direction,and leave a large of prediction residual at the edge of the cloud,a directional anisotropic bilateral filter background predictor is designed.The filter scale and filtering direction of the background predictor are adjusted adaptively using the local statistical information of the infrared image.The simulation results show that the target signal preserving and the gain of the target signal-to-clutter is improved respect to gradient weight bilateral filtering and two-dimensional minimum mean square error background prediction method.Due to shortage that the traditional trajectory prediction method is dependent on the target motion model,the local weighted polynomial nonparametric learning technique is applied.The generalized cross-validation method is used to adaptively update the bandwidth factor.The simulation results show that the method has better prediction ability and noise suppression effect than the least squares method.Aiming at the problem of poor dim target detection performance based on spatial domain and dim target detection based on temporal domain variance,cannot adapt to the variety of target motion,a multi-channel dim target detection strategy is proposed.Detecting the stationary target with the spatial channel,detecting the fast moving target with the short window temporal domain variance channel,detecting the slow moving target with the long window temporal domain variance channel,and integrating the multi-channel detection output with the maximum rate criterion.The simulation results show that this method has a high detection probability for different environments and different motion rate targets.