Application Of Deep Learning In Military Aircraft Recognition And Detection

Purpose—Automatic identification of military targets has always been an urgent problem in the military field.How to make a quick positioning and accurate recognition for warplane in the massive-scale image and video streaming and that makes a lot of sense.Thanks to the development of artificial intelligence,especially the advancement of target recognition and detection technology based on deep learning algorithms,making it possible to identify and detect aircraft in an intelligent way.For this purpose,this paper devoted itself to the study the deep learning methods to complete military aircraft recognition and detection.Method—In the image recognition stage of military aircraft,a convolutional neural network method was proposed to extract its features and determine the corresponding model.By building a convolutional neural network with an appropriate depth and using data enhancement methods to solve the problem of too small data sets.At the same time,in order to further improve the recognition accuracy of the convolutional neural network model,a series of optimization operations are performed on the model.In addition,use Verification set and test set to evaluate the recognition model ability.In the military aircraft target detection stage,Inception V2 convolutional neural network is used as a feature extractor,combined with Single Shot MultiBox Detector and Faster RCNN algorithm to train military aircraft target detection models,and the training results are compared.In the detection process,the data enhancement method is used to amplify the image data set,and the transfer learning method is used to fine-tune the target detection model.Finally,display and analyze the results of training,verification and testing.In the experimental stage,the deep learning framework TensorFlow was selected to complete the relevant experiments.Findings—In the stage of image recognition,it was concluded that the effect of 14-layer network training is better when comparing the training results of recognition models with different depths.Meanwhile,in order to improve the recognition accuracy of the model,we added batch normalization to the convolutional neural network.In addition,a cascaded rectified linear unit method was also been used to optimize CNN,it can further improve the recognition accuracy of CNN.In the military aircraft target detection phase,after comparing the training results of Faster R-CNN and SSD target detection algorithms,we found that the SSD algorithm is more suitable for detection.In the end,two different data format,that is,the images and videos streaming,were used to validate the model accuracy,and conclude that the SSD algorithm performs better detect effect on small targets.Limitations of research—The understanding of CNN is far from enough for the military aircraft target.The process remains to be further studied due to the lack of learning detailed features.Practical implications—Compared with traditional methods,applying deep learning to military aircraft image recognition and detection can not only further improve the accuracy,but also simplify the corresponding recognition and detection process to obtain a enhanced model.Moreover,if the model is used in the national airspace defense project in the future,for instance,using computerized modeling simulation,signal assistance together with model recognition will enhance combatants conduct enemy aircraft reconnaissance and enemy aircraft type identification accurately.Value—Proposed the use of deep learning technology to detect and identify military aircraft,which makes up for the shortcomings of traditional methods,that is,the cumbersome process and the low accuracy.Used CNN to extract and learn the features of military aircraft in the image,and a series of optimization methods to further improve the classification ability of CNN to military aircraft image.SSD target detection algorithm is more suitable for military aircraft target detection,not only positioning the target,but also identifying the type of military aircraft in the static image and dynamic video streaming.