A-CDM Milestone Event Time Acquisition Technology Based On Video Analysis

The goal of Airport Collaborative Decision-Making System(A-CDM)is to integrate airports,air traffic control,airlines and other civil aviation related units into a unified platform and share data by integrating data from all parties.Finally,the civil aviation communication system with airport as the communication center and unified and reasonable dispatch of resources of all parties will be realized.At present,some ground support operation data in ACDM system can not be automatically acquired.Aiming at this problem,this paper proposes a video analysis-based milestone event time acquisition technology for A-CDM system,which can solve the problem of difficult data acquisition such as the time of airport milestone events during flight transit.The research contents of this paper include determining the occurrence time of milestone events in A-CDM system based on context target detection and identifying the historical data of milestone events in A-CDM system based on prior knowledge.At present,most airports record the time of milestone events during flight transit by paper work order.This important data source can be well corrected by comparing the occurrence time of milestone events identified from the video,which improves the accuracy of the collected occurrence time of milestone events.Context-based target detection determines the time of A-CDM system milestone events through cameras distributed on the airport apron to identify aircraft,staff and ground support vehicles in the video.During the airport flight passing,the aircraft glides along the fixed taxiing curve at a relatively stable speed.Ground support operation has its own unique workflow,which ensures that the position of vehicles and aircraft is relatively fixed.Using the relationship between these contexts,the taxiing curve of an aircraft is fitted by the least square method to identify the specific position of the aircraft in motion.The background image is constructed by using continuous multi-frame image and inter-frame filtering method.The target image and background image are compared pixel by pixel,where the position of the moving object is greater than zero in the difference image,and the relative position and shape of the moving object are represented by the difference method.After obtaining the specific position of the aircraft in the video and the position and shape of the vehicle relative to the aircraft,the type of ground support vehicle is determined by the improved network model,and the time of milestone events is judged by combining the prior knowledge of the airport ground support process.The purpose of recognizing the historical data of milestone event occurrence time in A-CDM system based on prior knowledge is to automatically recognize the number of airport paper record sheets.The model adds a filter layer to the traditional recognition network to delete the results that do not conform to the characteristics of time data.The time difference probability density function of each milestone event is fitted by the mixture Gaussian model,and then the recognition results are selected by using the prior knowledge.Experiments show that both methods can effectively identify the occurrence time of milestone events,and the combination of the two methods can make the recognition accuracy of the occurrence time of milestone events better.