Target Detection And Location Based On The Panoramic Monitoring System

With the development of society,surveillance system plays an increasingly important role in modern society.A single-function monitoring system can no longer meet the needs of modern social development.How to realize the multifunctional monitoring system becomes the trend of research and development in the future.It has become a hot topic to realize the detection and location of the target with a large field of vision.Based on the research results of predecessors,this paper studies the fish-eye camera calibration algorithm,target detection algorithm and stereo matching algorithm to achieve the goal of target detection and localization.The structure and the imaging model of the fish-eye camera are studied.The original image collected by the fish-eye camera is processed with the methods of grayscale,histogram equalization and image smoothing.The principle of using the panoramic vision system to achieve target positioning is studied.The calculation process of plane calibration method is studied.The paper compares and analyzes various corner detection schemes in the calibration process.Proposes an improved corner detection algorithm based on subpixel point to deal with the problem of corner detection's accuracy is not high in the existing calibration method.The effect of the improved algorithm based on subpixel point corner detection is analyzed through experiments.An algorithm framework based on deep learning target detection is built.A feature extraction model based on multiple connection methods is proposed to solve the problem of large computational complexity in the existing deep learning algorithm.The model is optimized by optimization methods such as weight attenuation,data amplification,and batch normalization.Trains the data set of the target model and analyzes the results of test.The local stereo matching and global stereo matching algorithm are analyzed.An improved algorithm based on minimum spanning tree is proposed to solve the problem that the existing stereo matching algorithm can not balance real-time and precision.Compares and analyzes this algorithm with existing algorithms through experiments.A system based on binocular fish-eye lens detection and localization is set up.The logical structure of three algorithms of camera calibration,target detection and stereo matching are optimized.The simulation of sea state motion was carried out to verify the monitoring effect of target detection and location system.