Research On Airport Runway Intrusion Warning Technology Based On Video Processing

Runway intrusion refers to the situation that occurs in the airport and has adverse impact on the safety of the runway.On April 27,2006,the International Civil Aviation Organization(ICAO)gave the definition of runway intrusion,that is,all the wrong situations of aircraft,vehicles and pedestrians on the surface of the protection zone used for aircraft take-off and landing that occurred in the airport.For the possible accident consequences caused by runway intrusion,it is necessary to achieve real-time detection,rapid detection and timely warning.Due to the influence of light,weather,ground airport traffic conditions and target occlusion,all-weather real-time detection needs to consider efficiency and reliability.Combined with the advantages of deep learning in target feature learning,a method of airport runway intrusion warning prediction based on video processing is proposed.In this paper,the physical model of atmospheric scattering is introduced firstly,the principle of fog image degradation is described,and the fog image is de fogged.Aiming at the disadvantage of the dark channel prior de fogging algorithm,which is not real-time enough,through experimental demonstration,the original image is downsampled to reduce the calculation amount of transmittance,the maximum value of atmospheric light value is limited to increase the definition of contour,and the minimum value is filtered The filtering radius is set to increase the defogging effect and enhance the brightness of the image to avoid too dark.After the improvement of these parameters extraction methods,the time of image defogging is greatly reduced,too much CPU consumption is avoided,and the defogging effect is good,which saves time for target recognition.Then select the deep learning network yolov3 to detect the airport target in real time,identify the type and location of the target,configure the network training environment and parameters to train the network model of the airport target,and get the more accurate boundary frame according to the different clustering size of the target and different anchors.The generalization ability of the network model is evaluated,including the recognition ability of infrared image,the video target processing ability of infrared camera at night,and the generalization ability of the model is verified by the airport field test.Then,the airport runway intrusion warning system based on video processing is designed,which integrates defog algorithm,deep learning network and adopts Kalman filter to predict the movement state of airport target,which solves the problem of target trajectory prediction.In view of the problem that many airport targets are not easy to block,this paper proposes a detection method of shooting the airport at a large depression angle,and divides the airport into several regions to deal with the problem of absolute runway intrusion.At the same time,it judges whether there is a runway intrusion by calculation and takes alarm measures.At last,the system of airport runway intrusion warning based on video processing is realized,which is developed in Windows platform.At the same time,the experiment is carried out to simulate the airport environment.The model is used to simulate the intrusion behavior of various airport runways,and the system is used to judge it.The feasibility and stability of the system of airport runway intrusion warning based on video processing are verified.