Research On Intrusion Detection Of Unmanned Aerial Vehicle In Low Altitude Airspace

In recent years,with the gradual opening of low-altitude airspace,Unmanned Aerial Vehicle(UAV)have gradually expanded from the military field to the civilian sector.While it bringing various conveniences,it also led to frequent occurrences of ‘black flying' and ‘indiscriminate flying' incidents,which seriously endanger public safety and personal privacy.Therefore,the research on the intrusion detection of low-altitude airspace UAV is of great significance not only in public security control and safety in production but also anti-terrorism.In this paper,vision-based UAV detection technology is deeply studied.A method of UAV detection is proposed based on sky region.According to the motion state of the camera,it is divided into static scene UAV detection and dynamic scene UAV detection.The research content of this paper includes:?.UAV detection in static scenes: In the static scene,the inter-frame difference method,the optical flow method,and the background difference method are commonly used to detect moving targets.In this paper,according to the characteristics of flat gray area,large area and high brightness in the sky region,the image segmentation algorithm based on edge detection and the line scan method are used to segment the image to obtain the minimum circumscribed rectangle of the sky region.Then the UAV detection of the sky area in the rectangular frame is carried out by moving target detection algorithm.Based on the feasibility of the system,this paper uses Simulink to build a simulation system for experimental verification.The results show that the proposed method can effectively improve the accuracy of UAV detection and solve the problem of false detection caused by ground moving targets.?.Target detection of dynamic scenes: Dynamic scene target detection methods include background compensation,machine learning,and deep learning.In order to realize the intelligent monitoring of UAV,this paper firstly constructs the sky region segmentation network based on the full convolutional network,and realized the segmentation of the sky region pixel level.Then,the FastBox-based UAV detection network is constructed to realize the accuracy detection of UAV.The experimental results show that the sky region segmentation network and the UAV detection network based on deep learning can accurately realize the segmentation of the sky region and the detection of the UAV,and meet the real-time requirements.Finally,the fusion research of sky region and UAV detection network is carried out.By sharing convolution to extract features,and connecting the sky region and UAV detection network after the depth feature map,eventually in a model,the simultaneous detection of sky region and UAV is realized.