Performance evaluation for point target detection algorithm under infrared cloud clutter

Aiming at the limitations of current performance evaluation methods for point target detection under infrared cloud clutter, a new evaluation method is proposed. Firstly, for the elements that affect detection performance, the quantitative model of clutter, detector noise model, and quantitative and transfer model of target energy are constructed; in combination with the detector theory, the algorithm's performance characterization parameter is built; then, using quantified results of the above models as inputs, the mathematical relationship between algorithm's performance characterization parameter and quantification results is developed using back-propagation (BP) neural network on the basis of genetic algorithm; finally, taking Top-Hat algorithm and Butterworth filter as an example, the method is applied to evaluate and analyze its target detection performance. Experimental results show that the error of the method is below 5×10^-4, which shows high assessment accuracy. The proposed method uncouples the coupling factors influencing the detection performance for the first time. The proposed model is helpful for grasping the change law of the algorithm performance. The proposed model also provides the foundation for the general design of detection system and the selection of algorithms. It has theoretical significance and engineering application value.