Research On Laser Interference Effect Analysis And Evaluation Technology Of Photoelectric Imaging System

Laser is expected to be applied to the confrontation process of photoelectric imaging systems due to its monochromaticity,directionality and high brightness.Therefore,it is of great military value and practical significance to carry out research on the interference evaluation technology of laser to photoelectric imaging system.This paper starts from the perspective of laser interference image quality evaluation and the impact of laser interference on target detection.The specific research work can be summarized as follows:At present,most of the full reference laser interference image quality evaluation algorithms need to know the position information of the interference spot and the target in advance,so that the evaluation process is restricted by the prior knowledge and the preprocessing method.Aiming at this problem,this paper proposes a laser interference image quality evaluation algorithm based on convolution feature similarity(CNNSIM),which analyzes the output characteristics of the image before and after laser interference in the convolution network,and utilizes the hierarchical and occlusion of features.The sensitivity of the occlusion of the key information in the interference image is evaluated,and the input requirement of the target/spot position information is avoided.The simulation experiment verifies the effectiveness of the new evaluation algorithm in different scenarios.Aiming at the problem that the reference image is difficult to obtain in practical applications,this paper starts with the prediction processing of the occluded information,and improves the Markov Random Field Estimation Algorithm(MRF),realizing the real-time estimation of the occlusion area information;Based on the statistical characteristics of laser interference images in natural scenes,a non-reference evaluation algorithm(NSSIE)based on natural scene statistics and occlusion region information estimation is proposed.Compared with the traditional algorithm,the new algorithm does not need a reference image and can accurately reflect the quality loss of the laser interference image.Large-scale simulation experiments verify the effectiveness of the new evaluation algorithm.On the other hand,this paper also studies the interference effect analysis method from the perspective of the impact of target detection.Firstly,the effects of laser interference on Faster-RCNN and YOLO-V3 detection algorithms are compared and analyzed from the two aspects of target occlusion rate and target similarity.Then,a target-oriented interference spot effective interference region partitioning method(TOEDZ)is proposed.The method uses the simulated image,target and template similarity,and target detection algorithm to determine the effective interference similarity threshold of the target under the detection algorithm.It was migrated to the interference effective region of other spot images,and large-scale simulation experiments verified the effectiveness of the partition.