视觉SLAM中图像特征点提取与匹配算法研究

Research on feature extraction and matching algorithms for visual simultaneous localization and mapping

在视觉SLAM(visual Simultaneous Localization And Mapping,vSLAM)中,利用提取到的图像特征点进行相机位姿估测是一种非常重要的位姿估算方法,为了实现相机的定位,图像特征点必须具备鲁棒性、尺度性和高效率等特点。介绍了图像特征点提取与匹配在视觉SLAM中的作用和场景中图像特征点需要具备的特性;对几种主流的图像特征点提取算法,即尺度不变特征变换(Scale Invariant Feature Transform,SIFT)算法、加速稳健特征(Speeded-Up Robust Features,SURF)算法、快速特征点提取与描述(Oriented FAST and Rotated BRIEF,ORB)算法分别进行了简要说明;并通过设计实验,在室内环境中,对几种图像特征点提取算法的运行效率和图像特征点匹配正确率进行了对比测试。实验表明,ORB算法在运行效率和匹配正确率上占据优势,能够较好地满足视觉SLAM中实时性和鲁棒性的要求。

In visual SLAM(visual Simultaneous Localization And Mapping,vSLAM),it is a very important method to estimate the pose of the camera by using the extracted image feature.In order to locate the camera,the image feature must have the characteristics of robustness,scalability and high efficiency.The role of feature extraction and matching in visual SLAM and the characteristics of feature in the scene is introduced.Several main feature extraction algorithms,Scale Invariant Feature Transform (SIFT), Speeded-Up Robust Features (SURF) and Oriented FAST and Rotated BRIEF (ORB),are briefly introduced.The efficiency and matching accuracy of the feature extraction algorithms were compared and tested by designing experiment.The result show that ORB algorithm has advantages in running efficiency and correct matching rate,and can meet the requirements of real-time and robustness in visual SLAM.