Military Target Detection And Threat Assessment In Complex Environment

In modern warfare,the increasing number of unmanned intelligent equipment,the automatic identification of military targets in complex environment and the assessment of threat degree make one of its core technologies.Based on the military targets in complex environment,combining the techniques of target feature extraction,convolution neural network,anchor frame selection and non-maximum value suppression,this paper carries out the study of real-time rapid detection technology of visible light military target images,and on this basis,the threat level assessment of military targets is mainly carried out using fuzzy set analysis theory.This paper uses the Internet to collect images of military targets in real-world environments,analyzes the characteristics of military targets in complex environments,references the PASCAL VOC data set format,and uses artificial marking methods to build data sets of military targets including tanks,missile launchers,soldiers,military vehicles,fighter jets,helicopter gunships,fixed-wing drones,rotary-wing drones,surface ships and other military targets.Based on the Military-Data data set,the Faster R-CNN algorithm with fast detection speed and high precision is selected as the basic algorithm for military target detection,and then the size scale setting of the target anchor frame is improved by reference to clustering,the occurrence of target leakage and mis examination is reduced,and the accuracy of high retrieval rate is improved by using Soft-NMS instead of NMS(non-maximum value suppression).The target training in the Military-Data data set obtains the neural network detection parameters,makes the target overall recognition accuracy reach 92%,and finally uses the UVA to collect the tank in the real scene,verifies the accuracy and generalization of the algorithm,and the test and laboratory simulation results are consistent.In view of the military target test results,the four factors of target type,target mobility,target distance and target attack are selected as threat factors,first the target membership function is obtained by fuzzy set,the second is to use the information entropy to obtain the weight of the target attribute,and finally the target threat degree is sorted by the TOPSIS theory,and the target threat degree is sorted by the TOPSIS theory.The above research achieves the real-time and accurate detection and threat assessment analysis of military targets in complex environment,and can provide technical reference for intelligent unmanned weapons and equipment in the future.