Application Research Of R-CNN-based Object Detection In Preventing Runway Incursion

The runway serves as the starting and ending facility for aircrafts operation,and its safety is the most important part in the safety of civil aviation.With the rapid increase of flight flow,the problem of runway incursion has become a safety problem that must be solved.The human defense alone cannot effectively prevent runway incursions.There is an urgent need to prevent runway incursion through technical measures.Aiming at the problem of automation of runway intrusion prevention,this paper proposes a system for preventing runway incursion based on deep learning image object recognition algorithm,focusing on the accuracy of deep learning object recognition algorithm for target object recognition in the image,as well as the problem of airport camera layout,And the logical design to judge runway incursions based on the detection results.The focus of the research work in this article is as follows.First,the image detection algorithms based on deep learning are compared.According to the comparison results,the latest Mask R-CNN algorithm is selected as the core detection algorithm of this article.The algorithm principle of Mask R-CNN is explained in detail to improve the specificity of the algorithm.Recognition accuracy and real-time,based on the weights obtained by the algorithm training,using transfer learning to train the self-built airport target data set,which improves the detection accuracy of the algorithm for specific airport targets.Secondly,the specific layout method and selection basis of target detection hardware equipment are proposed.Based on the parameters of different runways in the airport,the layout method of the camera is designed.On the basis of ensuring the accuracy of target recognition,the detection results are summarized and analyzed to determine the aircraft.Relative position,a digital logic circuit is designed to complete the determination of runway incursions.Finally,we built a hardware and software platform for the laboratory to verify the effectiveness of the system,and designed and built a scaled-down airport sand table model based on the actual airport operating environment.The channel for data transmission between devices is designed to complete the verification of the real-time,accuracy and effectiveness of the system.Experimental results show that the accuracy of detection of runway incursions can reach more than 90%.