Research On Aircraft Wake Detection,Identification And Intensity Classification Based On Deep Learning

Aircraft wake is a pair of strong vortices with reverse rotation formed behind the aircraft during flight.It has the characteristics of large spatial scale(100 meters to kilometers),fast rotation speed,stability and long existence time(minutes or even more than ten minutes).When the rear aircraft encounters the wake of the front aircraft,it may produce jitter and turbulence,which will seriously cause the aircraft to lose control and cause heavy losses.It is urgent to develop the real-time detection technology of aircraft wake.This paper mainly studies the problem of lidar detecting aircraft wake in clear sky.The deep learning method is applied to the detection data to quickly identify,detect and grade the intensity of aircraft wake,so as to provide technical support for military and civil aviation aircraft take-off and landing safety control.In the first aspect,an aircraft wake identification method based on alexnet network model is proposed.Based on the lidar detection data,the alexnet network model is improved according to the aircraft wake characteristics and identification requirements,and then the model is trained based on a large number of wake observation data.The test results show that the model can identify the aircraft wake over the airport runway with high accuracy,and can effectively provide reference for the wake hazard early warning of air traffic control.In the second aspect,an aircraft wake detection technology combining lidar detection and yolov5s network is proposed.This method optimizes the yolov5s network model according to the characteristics of aircraft wake.Aiming at the shortcomings of the original model loss function,ciou is used＿Loss as a new loss function.Then,based on a large number of lidar aircraft wake observation data,the improved yolov5s model is trained,and then the position and characteristic information of aircraft wake are extracted to detect aircraft wake targets.After testing,the detection accuracy of the optimized model is improved by 2.56%,and the aircraft wake target can be detected with high accuracy and speed.In the third aspect,a classification method of aircraft wake intensity based on yolov5s network is proposed.The velocity loop of aircraft wake is an important characteristic parameter to characterize the wake intensity.The larger the loop,the greater the wake intensity.Firstly,the path integral algorithm is used to calculate the velocity circulation of wake data.Then,referring to the circulation based recat-new model classification standard,the wake data set is divided into three categories: low intensity,medium intensity and high intensity.Using the prepared data set,the yolov5s network model is trained,and then the characteristics of aircraft wake with different intensity are extracted to realize wake intensity classification,and the application of wake classification method in engineering is analyzed.Using lidar to detect the front aircraft wake in real time and using the wake intensity classification method based on deep learning to accurately detect the front aircraft wake and classify its intensity is of great practical significance to improve the aircraft take-off and landing efficiency,ensure flight safety and improve the combat effectiveness of the army.This paper proposes to apply the deep learning method to the rapid detection,identification and intensity classification of aircraft wake,closely follow the real-time detection requirements of aircraft wake,and carry out the training of aircraft wake identification and detection model.The field measured data verification shows that the performance of relevant indexes of the model is excellent.The research results provide technical support for the takeoff and landing safety control of military and civil aviation aircraft.