Research On The Algorithm And System Realization Of The Discovery And Tracking Of Moving Targets In Dynamic Scenes

Target tracking is a very important branch of image processing problems.Due to the influence of various factors,such as illumination,motion blur,target occlusion and rotation during the tracking process,target tracking becomes one of the difficulties in image processing.Meanwhile,in the dynamic scene,the discovery and tracking of moving targets put forward higher requirements for the tracking algorithm.The dynamic scene means that the camera is mounted on the motion platform.Due to the movement of the photographing platform during the capture of the target image,the background in the image also moves.This makes the tracking process more difficult.In this paper,through continuous and in-depth study of the existing tracking algorithms,a set of moving target tracking system based on monocular camera and pan-tilt is designed,and the following research work is carried out around the system:(1)Design a tracking and tracking system for moving objects in dynamic scenes by researching related filter tracking algorithms.The system has a simple hardware structure and low cost,making it relatively easy to implement system products.(2)By studying several mainstream image feature extraction algorithms,image feature extraction experiments and feature matching experiments are designed to compare the effects of SIFT(Scale Invariant Feature Transform)features,ORB(Oriented FAST and Rotated BRIEF)features,and their respective characteristics.The experiment mainly compares the number of feature points extracted by the two algorithms and the running speed,and then through matching experiments to observe the detection effect of the target in the image.(3)Through the study of several correlated filter tracking algorithms,according to the shortcomings of these algorithms in long-term tracking problems,this paper proposes a correlated filter tracking algorithm that combines target detectors.The algorithm introduces a target detector to avoid the problem of tracking objects being replaced by occluders due to occlusion of the target.(4)The target tracking experiment was designed to verify the accuracy of this algorithm,and compared with other tracking algorithms on this basis.Several experimental scenarios are designed to verify the robustness of the algorithm.The experimental results basically meet the requirements.