Moving Target Detection In Remote Tower

As a new technology with low cost,high efficiency and remote control for airport,remote tower is of great significance to reduce the operating cost of airport and ensure the safety of airport operation.In recent years,attention has been paid to the application of remote tower technology at home and abroad.The moving target detection of remote tower is the core of its technology.It is also the premise of aircraft target recognition,conflict behavior prediction and other functions.However,the detection targets of remote tower are mostly remote scenes and small targets.And there are many moving objects in airport scenes.Therefore,it is easy to encounter occlusion problems.In view of the above problems,this paper has carried out in-depth research on the detection algorithm of moving targets of remote towers,which mainly includes the following aspects:Firstly,the traditional algorithm of moving target detection and the theory of moving target detection algorithm based on deep learning are studied,and the advantages and disadvantages of each algorithm in the application of remote tower are emphatically analyzed.Secondly,this paper introduced the concept of Repulsion Loss(RL)repulsion loss,improved the loss function of the current You Only Look Once(YOLO)target detection algorithm,A target detection algorithm RL-YOLO based on RL loss function is proposed.It can be seen through experimental verification that the moving target detection algorithm with improved loss function improves the accuracy of target detection and solves the occlusion problem in moving target detection.By analyzing the pyramid structure of Feature Pyramid Networks(FPN),referring to the application of FPN in Regional Proposal Network(RPN)and the network structure of Single Shot MultiBox Detector(SSD),a network architecture FPN-YOLO which combined FPN with YOLO was proposed.The structure can detect moving targets in different scale images,and improve the precision of small target detection.Finally,a data fusion target recognition algorithm based on ADS-B was proposed,which realized the fusion of combining the remote tower target detection data with ADS-B aircraft data and matching the sign.Combined with the reality video and ADS-B monitoring data acquired by a training airport,the sign matching algorithm was verified.The related algorithm research of this thesis has applied preliminary research on airport target detection and ADS-B target matching,which laid a technical foundation for the practical application of remote tower and further target abnormal behavior detection.