Research And Implementation Of Tey Technology Of Multi-Target Tracking Based On DSP Image Processing System

In the process of scientific and technological development,the traditional manual operation method is far from meeting the requirements of modern enterprises,and production automation is an inevitable trend.Under this demand,artificial intelligence technology has also been strongly developed.Image recognition and processing technology is a special application direction of artificial intelligence technology.With the growing demand for various types of video intelligent analysis,the importance of using excellent models and artificial intelligence technology for image recognition and video analysis has been paid more and more attention.Image processing technology is one of the basic capabilities of intelligent video analysis,and it is also the focus of research by many scholars.In this paper,the multi-target tracking technology in image and video processing is deeply studied,and based on the detection and recognition algorithm of image targets,a theoretical scheme of multi-target tracking is proposed.Finally,the relevant research algorithms are applied to the image processing system based on DSP chips,and at the same time,various scenarios are simulated,the simulation environment of the image processing system is constructed,and experiments and data analysis are carried out.The main research contents and innovative work of this paper are as follows:(1)Aiming at the problem of object recognition in substations,this paper studies common image target recognition algorithms.Based on the image target recognition algorithm based on optical flow method combined with feature detection,the detection effect is not ideal when the light background is not good.(2)Based on the image target recognition algorithm,a multi-target tracking implementation scheme is designed.The overall technical route of "identification first,tracking later" is adopted,that is,multi-dimensional correlation is carried out through the target data output by the identification algorithm,so as to achieve the tracking of various targets.In this paper,a theoretical model and scheme of multi-target tracking based on preprocessing and time window are designed,and a trajectory fusion strategy under this scheme is proposed to achieve real-time tracking of multiple targets.Compared with the conventional method,the trajectory fusion strategy under this time window adds a post-error correction mechanism,which can correct the errors and deviations caused by the trajectory processing of the previous window.At the same time,in the trajectory fusion strategy,this paper uses 3 consecutive windows in time for comparison,and the conventional 2 kinds of result judgments are optimized to 5 kinds of result judgments,which can effectively reduce the accumulation of errors.(3)In order to verify the feasibility and accuracy of the relevant algorithms and implementation schemes,in the research process,for the application of the relevant algorithms and schemes,a scheme for multi-target tracking in an image processing system based on DSP chips was designed.The image processing system is operated and executed,and experiments are carried out.In the experimental environment,compared with the traditional multi-target tracking technology route,the average window response time is 78.6ms,which is 7.5%shorter than the conventional DSP response time of 85 ms,which meets the actual use requirements.