基于立体视觉的目标检测与轨迹预测研究综述  

Survey of Target Detection and Trajectory Prediction Based on Stereo Vision

作　　者：胡春生[1] 闫小鹏 魏红星 李国利 HU Chunsheng;YAN Xiaopeng;WEI Hongxing;LI Guoli(College of Mechanical Engineering,Ningxia University,Yinchuan 750021,China)

机构地区：[1]宁夏大学机械工程学院,银川750021

出　　处：《计算机工程与应用》2022年第3期50-65,共16页Computer Engineering and Applications

基　　金：宁夏大学研究生创新项目(GIP2021008)。

摘　　要：目前基于立体视觉信息的运动目标识别定位、跟踪及轨迹预测是机器视觉领域的研究热点。通过归纳整理相关文献,从双目立体视觉技术、运动目标检测技术、运动目标轨迹预测技术三个方面对基于立体视觉的运动目标检测及轨迹预测进行了概述,分别阐述了相机标定的常见方法、图像特征提取及立体匹配不同算法的适用场景、各运动目标检测方法的优缺点、常用轨迹预测方法的预测思路及优劣势。针对基于立体视觉的运动目标检测与轨迹预测的研究,指出了其现今面临的挑战和研究重点,可为相关研究人员提供参考。At present,the recognition,positioning,tracking and trajectory prediction of moving targets based on stereo vision information are research hotspots in the field of machine vision.This paper summarizes the moving target detection and trajectory prediction based on stereo vision from three aspects of stereo vision technology,moving target detection technology and moving target trajectory prediction technology by summarizing relevant literature.The common methods of camera calibration,the applicable scenarios of different algorithms for image feature extraction and stereo matching,the advantages and disadvantages of each moving target detection method,and the prediction ideas and advantages and disadvantages of common trajectory prediction methods are respectively described.Finally,the paper points out the challenges and research focus for the research of moving target detection and trajectory prediction based on stereo vision,which can provide reference for related researchers.

关 键 词：立体视觉 相机标定 图像特征提取 立体匹配 运动目标检测 轨迹预测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]