单目视觉室内无人飞行器运动目标测距技术研究  

Research on Moving Targets Distance Measurement Technology of Indoor UAV with Monocular Vision

作　　者：邓孝逸[1] 曾庆化[1] 徐贵力[1] 朱华[1] 

机构地区：[1]南京航空航天大学自动化学院,江苏南京210016

出　　处：《计算机与现代化》2013年第3期152-155,164,共5页Computer and Modernization

基　　金：国家自然科学基金资助项目(60904091&61104188);国家重点基础研究发展计划(973计划)(2009CB724002);航空科学基金资助项目(20100112005);江苏省高校优势学科建设工程项目;中央高校基本科研业务费专项基金资助项目(NP2012103&NP2011049)

摘　　要：为了准确检测动态背景中运动目标并对其测距,实现室内无人飞行器避障需求,本文研究一种基于单目视觉的运动目标测距方法。在惯性信息辅助提高运动目标检测跟踪效率的同时,推导基于小孔成像几何约束的测距算法。在初步解算距离情况下,研究针对摄像机俯仰角误差的测距补偿方法。实验结果表明,本文算法不仅能够完整检测和跟踪运动目标,并且便于准确快速求取运动目标到摄像机的距离,具有重要的工程参考价值。In order to detect moving targets accurately under dynamic background and achieve its distance measurement to meet the requirements of the indoor UAV obstacle avoidance,a moving target distance detection method based on monocular vision is presented.The inertial information is used to accelerate and improve the efficiency of detecting and tracking moving targets,distance measuring algorithm that derived from geometric constraints based on Pinhole imaging.The angle error of the camera is corrected on the case of preliminary measured values.Experiments show that the research in the paper can not only detect and track a moving target completely but also calculate the distance from the moving target to the camera quickly and accurately,which has significant engineering reference value.

关 键 词：单目视觉 动态背景 惯性信息辅助 目标检测 实时测距 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]