| In recent years,with the increasing application of UAV in aerial photography,express delivery,survey,search,plant protection and other aspects,people’s quality of life and social efficiency have been rapidly improved.However,with the rapid development of UAV market,the unlicensed flight and random flight of UAV bring serious threats and troubles to national security and social stability.How to counteract the threat of "black flying" UAVs is a n urgent problem to be solved.The UAV threat countermeasure system is an anti-UAV system.Image intelligent recognition is one of the most effective methods to recognize UAV s.How to use image recognition method in anti UAV system to identify UAV more acc urately has become the focus and difficulty of research.The subject of "C-Drone System Technology Research and Development" is a sub-project of the key science and technology research and development project of Jilin Province Science and Technology Depart ment.It mainly studies the problem of intelligent image recognition in anti-UAV systems,and proposes an intelligent recognition method with higher recognition accuracy.Firstly,an image recognition method of anti-UAV system based on a support vector machine is proposed.The model based on SVM is designed.The image recognition UAV is realized by using SIFT feature.Secondly,an anti-UAV system image recognition method based on a traditional convolutional neural network is proposed.A model based on traditional convolutional neural network is designed.Aiming at the over fitting problem in the training process,the network is optimized to realize the image recognition of UAV.Then,an improved convolutional neural network image recognition method for anti-UAV systems is proposed.A convolutional neural network model based on multi-layer feature fusion is designed,and the features of different convolution layers are fused to realize the image recognition of UAV.Finally,the recognition experiments of three image intelligent recognition methods show that the accuracy of SVM is 89.4%,the accuracy of traditional convolution neural networks is 96.2%,and the accuracy of multi-layer feature fusion convolution neural networks is 98.1%.The experimental results show that the recognition effect based on the multi-layer feature fusion convolutional neural network is better than the other two methods,and it can be used for the image recognition of anti-UAV systems,providing a reference for similar research. |
