Anti-jamming Recognition Algorithm of Infrared Target Based on Machine Vision
In complex air combat,the number of artificial decoys is large,which makes it easy to cause occlusion and stickiness to the actual target,resulting in low accuracy and poor stability of infrared target detection.In order to solve the above problems,the feature fusion algorithm and the adaptive Yolov3 target detection algorithm are organi-cally fused,and the target extraction rate is improved by shape similarity,and finally,the DFE-YOL-3 infrared target anti-jamming detection model is constructed.In this model,the median filtering algorithm and histogram equalization algorithm are used to denoise the original image to improve the recognition of the image target,and then the gray co-occurrence G feature and histogram H feature of the infrared image are extracted and fused respectively to construct the fusion feature vector,which improves the feasibility of image detection.Then the DBSACAN shape clustering algo-rithm is used to improve the detection accuracy of infrared small targets,and the IOU threshold is adaptively calculat-ed based on the fusion features.Finally,the target regression box is constructed by optimizing the loss function of the infrared target position.The simulation results of the experimental model show that the accuracy of the DFE-YOL-3 algorithm is the highest,up to 94.38％,with an average increase of 5.68％ compared with other traditional infrared anti-jamming detection algorithms on the IACD infrared air combat simulation data.The recall rate of the DFE-YOL-3 algorithm is also the highest,which is 93.62％,and the average increase is 4.10％,that is to say,the DFE-YOL-3 algorithm has high accuracy and stability.At the same time,the DFE-YOL-3 algorithm has good modeling timeliness and detection timeliness.To sum up,the DFE-YOL-3 algorithm solves the problem of artificial decoy oc-clusion and stickiness,effectively improves the accuracy and stability of infrared target anti-jamming detection,and has a certain simulation application value.
NETL
NSTL
万方数据
