Research On Information Fusion Method Of Multispectral Compounding Detection

Target detection and recognition are the key techniques of intelligent weapon system under complex environment.In the real battlefield environment,single detection system can no longer meet the target detection requirements under strong interference conditions,and the composite detection system has become a research hotspot in the field of near-field detection.Currently in the military field,laser detection technology and infrared detection technology are widely used in battlefield target detection and recognition.By combining these two photoelectric detection systems,it can effectively solve the shortcomings of insufficient laser anti-smoke ability and infrared detection to be easily interfered by the background.It can effectively improve the reliability of target recognition in the battlefield.In this paper,the 32-element linear array laser/infrared composite scanning imaging fuze is taken as the research object.Aiming at the problem of fusion identification of armored targets in the near-field detection condition,the imaging mechanism and imaging characteristics of linear laser/infrared composite scanning are analyzed,multi-band information characteristics of the target are studied,and the multi-band information spatiotemporal fusion algorithm based on image frame correlation is proposed.In the first,the near-field transmission characteristics of the laser and target thermal radiation characteristics are analyzed.determined the working band of laser and infrared.Combined with the principle of push-scan scanning imaging,the linear array laser/infrared composite scanning imaging model was established.The frame parameters and image resolution of linear laser/infrared composite detection fuze imaging are studied.Then the top contour features and thermal radiation characteristics of typical targets are extracted,based on these two characteristics,two BP neural networks are constructed,and the preliminary discrimination results of laser and infrared detection methods are obtained.According to the characteristics of different information fusion levels under near-field detection conditions are analyzed,the information spatio-temporal fusion strategy of multi-band composite detection is proposed,and the multi-band information spatial domain fusion algorithm based on D-S evidence theory is studied.By correlating the targets between two frames of images,the multi-band information is temporally fused.Finally,based on the multi-band information spatio-temporal fusion algorithm of image-frame correlation,the final fusion recognition result of the target in laser/infrared composite detection is obtained.Through experiment,the detection data of linear array laser and infrared to armor target were obtained,and the frame image containing the target information was obtained by simulating the true scanning process of the fuze in the battlefield.The simulation results of the multi-band information spatial-temporal fusion algorithm based on inter-frame correlation show that the correct target recognition rate is about 90.0%,which is approximately 10%higher than the average target recognition rate of the multi-band information spatial domain fusion method,the validity of the proposed algorithm is verified.