摘要
针对移动机器人视觉SLAM过程中的关键环节图像匹配及相机位姿求解问题,通过ORB特征匹配算法匹配移动机器人采集的两帧图像,得到特征点集。分析了最小二乘法理论特征,根据两帧图像中特征点不变性,采用基于最小二乘法的ICP算法建立相机运动模型,使用SVD分解法求解运动模型,估计相机位姿状态,并通过非线性优化算法验证SVD分解法。利用RGB-D相机Kinect采集图像,对上述算法进行实验,得到准确的相机姿态数据,从而有效解决了三维空间下移动机器人局部移动过程中的定位问题。
In order to solve the key problem of image matching and camera pose in mobile robot visual SLAM process,it proposes ORB feature matching algorithm and matches the two frames captured from the mobile robot,obtains a set of feature points. It analyses the least squares theory characteristics,uses ICP method to establish the camera motion model according to the feature points of two frame image invariance,applies SVD camera pose estimation to build the motion model. It verifies the SVD decomposition method based on the nonlinear optimization algorithm,uses the RGB-D Kinect camera to capture images,tests the algorithm. The obtained camera pose data solve effectively the problem of three-dimensional positioning of mobile robot under local mobile process.
出处
《机械设计与制造工程》
2017年第11期13-18,共6页
Machine Design and Manufacturing Engineering
