| When tracking a single aerial target in the air,in order to evaluate the current motion state of the target,it is necessary to increase the focal length of the camera for attitude judgment.However,due to the characteristics of high speed and large scale change of aerial targets,blindly increasing the focal length will cause the target to be easily out of the field of view,so that effective detection and tracking cannot be carried out.In addition,when the aerial target is blocked by obstacles,the camera with the increased focal length is more likely to produce the problem of "blind",which further causes the target to disappear from the field of view and cannot complete the detection and tracking task normally.In order to solve these problems,taking common air target aircraft as an example,this paper presents a dual-field of view air target detection and tracking method.The main research work of this paper is as follows:(1)Overall Design of Dual-Field of view Target Detection and Tracking System.To address the issues encountered during the detection and tracking of aerial targets,a dual-field of view airborne target detection and tracking method is proposed.A tracking and monitoring system that supports this method is designed,and a hardware analysis of the entire system is conducted.The selection and design of the experimental platform for this system and the simulation of the motion model are completed;(2)Research on Large Field of View Motion Target Detection Algorithm.Aiming at the problem that the moving target detection algorithm in the static background is not applicable due to the dynamic background interference generated by the camera movement when the dual-field of view aerial target detection and tracking system is shooting with the large field of view camera,an improved moving target detection algorithm based on dynamic background compensation is proposed.This algorithm avoids false and incorrect detections caused by the partial motion of the video background and greatly reduces the time required for detecting airborne targets;(3)Research on Object Recognition Algorithm for Small Field of View.In view of the problem that the dual-field of view aerial target detection and tracking system requires a small field of view for identifying and analyzing aerial targets,we propose a design and construction idea for a small field of view object detection algorithm using the Yolov7-tiny algorithm.The creation of a dataset suitable for this project was completed,solving the problem of the lack of suitable datasets.During training,methods such as K-Means clustering and transfer learning were used to train the model,making the algorithm more accurate in detecting targets;(4)Design and Application of Dual-Field of View Aerial Target Tracking Algorithm.To solve the problem that the KCF tracking algorithm fails to track fast-moving or occluded targets due to inappropriate sampling and model updating,a tracking fusion algorithm based on KCF and Yolo V7-tiny is proposed.In addition,the application of the tracking algorithm is also studied.By real-time updating the offset distance and azimuth of the target position within the field of view during the tracking process,the PID control of the tracking holder is achieved. |
