【作者】 周玉林；

【导师】 李江昊；

【作者基本信息】 燕山大学 ， 光学工程， 2018， 硕士

【摘要】 随着人工智能时代的到来,移动机器人被广泛应用到生活和生产活动的各个方面。如何提高机器人目标的检测、识别和跟踪的准确性成为当前研究的热点。本文研究静态背景下的微小型机器人的识别与跟踪,在此基础上,又研究了动态背景下的微小型机器人的识别与跟踪。首先,本文分析静态背景下光流法,混合高斯模型的背景差分法和帧间差分法三种目标检测算法。动态背景下的基于图像块匹配的全局运动补偿算法。其次,在检测出运动目标的基础上完成对微小型机器人的目标识别。详细分析了基于ORB特征点匹配识别的算法原理。在微小型机器人的匹配特征点包含误匹配点的情况下,详细分析了RANSAC误匹配去除算法。针对该算法的缺点,采用基于空间一致性准则的RANSAC误匹配去除算法。然后,对运动目标进行跟踪。本文主要采用基于ORB特征点与卡尔曼滤波器相融合的运动目标跟踪算法。在跟踪发生错误的情况下,Kalman滤波器可以完成对微小型机器人的位置修正,保证跟踪的正确性。为了提高微小型机器人跟踪的实时性,本文提出了一种基于摄像机帧采集率和图像边缘检测算法相结合的图像关键帧抽取算法。最后,构建上位机视频监控平台。平台使用微软的MFC界面开发工具,开发了操作简单的上位机监控界面,在此界面显示微型机器人的运动情况。并且通过实验验证了微小型机器人的目标检测、目标识别与目标跟踪算法的鲁棒性与实时性。更多还原

【Abstract】 With the advent of the era of artificial intelligence,mobile robots have been widely used in all aspects of life and production activities.How to improve the accuracy of detection,identification and tracking of robot targets becomes the hotspot of current research.This paper studies the identification and tracking of micro mobile robots in static background.On this basis,the recognition and tracking of micro-robots in dynamic background are also studied.First of all,this paper analyzes the optical flow method in the static background,the background difference method of mixed Gaussian model and the three target detection algorithms of frame difference method.Global motion compensation algorithm based on image block matching in dynamic background.Secondly,the target recognition of the micro-small robot is completed on the basis of detecting the moving target.Detailed analysis of the algorithm principle based on ORB feature point matching recognition.When the matching feature points of micro-robots contain mis-matching points,the RANSAC mis-matching algorithm is analyzed in detail.Aiming at the shortcomings of this algorithm,the RANSAC mismatch matching algorithm based on the spatial consistency criterion is adopted.Then track the moving target,this paper mainly adopts a moving target tracking algorithm based on ORB feature points and Kalman filter.In the case of tracking error,Kalman filter can complete the position correction of micro-small robot to ensure the correctness of the tracking.In order to improve the real-time tracking of micro-mobile robots,Finally,the upper computer video monitoring platform was constructed.The platform uses Microsoft’s MFC interface development tools to develop a simple PC monitoring interface.In this interface shows the movement of the micro-robot.And through experiments,the robustness and real-time performance of the detection,recognition and tracking algorithms for micro-robots are verified.This paper proposes an image key frame extraction algorithm based on camera frame acquisition rate and image edge detection algorithm.更多还原