Research On Small Multi-rotor UAV Intrusion Detection And Interception System

In recent years,the country has promoted the rapid development of the entire drone industry through a series of policies and measures.The civil drone market has been expanding and the industry applications have been expanding,showing a vigorous development momentum,and has been very widely used in aerial photography,agricultural plant protection,disaster relief and rescue,terrain exploration,light shows and other aspects.However,the rapid development of the civil drone industry to bring convenience to people’s lives at the same time,but also to the public social security has brought great hidden dangers,a variety of illegal and illegal events,"black flight","indiscriminate flying" resulting in the impact of social security and stability of events There are incidents that affect social security and stability.It is extremely urgent to conduct research on "antidrone technology" for civilian UAVs.Based on the research of UAV image recognition algorithm,combined with UAV detection and countermeasure technology,an intelligent UAV intrusion detection and interception system is designed in this paper,which can play an important role in the security field and has wide application value:(1)In-depth analysis of various image characteristics and image classification methods,the HOG(Histogram of oriented gradients)feature extraction algorithm combined with nonlinear SVM(Support Vector Machine)UAV image recognition scheme is used for the characteristics of multirotor UAV targets,and the optimal parameters are determined based on the analysis and comparison of experiments The HOG+SVM recognition scheme under the experiment.To address the problem of slow recognition speed of the above recognition scheme,we propose a recognition scheme by adding Vi Be(Visual Background Extractor)motion target detection algorithm,which extracts the sub-images of the motion target region first and then classifies the sub-images for recognition,which will effectively reduce the computation of HOG+SVM recognition scheme and thus improve the recognition speed.(2)To address the problems of complex models,long training time,and low network generalization in traditional networks,this study proposes an improved Res Net-50 model based on existing research,i.e.,adding CBAM(Convolutional Block Attention Module)module to the residual structure and using CSPNet(Cross Stage Paritial Network)to improve the structure of the original network.The improved model is compared with the traditional deep learning model for experiments,and the results show that the improved model achieves better results in both UAV image classification and UAV target detection experiments,and has good prospects for engineering applications.(3)For the UAV intrusion problem in real scenarios,a tracking servo system based on PID control is designed,and based on this system combined with software algorithms such as UAV target detection and UAV countermeasure equipment,an intelligent UAV intrusion detection and interception system that can perform follow-on directional jamming is built to achieve accurate identification and tracking and interception of intruding UAVs.The results show that the system can automatically identify and track UAVs entering the target airspace,and drive the tracking servo system to track,target and interfere with the target UAVs in real time to achieve the purpose of intercepting the invading target UAVs.