用Kinect定位平面上单一圆柱型物体

通过Kinect设备获取物体所在场景的3D点云图像，再用积分图像计算场景中点云的法向量，根据法向量Y方向分量与XOZ平面所成角度的正切阚值滤除放置物体的平面，同时利用坐标输出阏值的设定滤除其它背景点和噪点。随机采取物体点云图中任意法向量不同的两点，根据它们的法向量和坐标，用异面直线公垂线的计算方法得到圆柱体的中心轴。再用点到空间直线的距离公式计算圆柱型物体的半径。最后通过计算物体点云X，Y，Z方向坐标的取值范围，实现对物体的定位。使用采集的场景点云图对算法结构进行测试，结果表明，算法结构可以可靠提取放置物体的平面和平面上的圆柱形物体，可以为机器人抓取提供准确数据信息。

Single cylinder object on planar location using Kinect

Using Kinect equipment obtains 3D point clouds of scene which contains the object, then computing the normal vectors of point clouds in the scene by integral pictures and filtering the closest planar with object.To filter the incompatible scene and noise points, we should output threshold value setting by the coordinate of point cloud.Randomly obtain the two points with different normal vectors in parts of the objective point clouds, bY their normal vectors and coordinate, make use of the calculation method of crossover line to find the central axis of the cylinder, using the distance between the point and spatial line to get the radius of the cylinder object. It can be positioned by calculated the value area for the objective point clouds in X, Y, Z coordinates, then testing the algorithm structure by the scene point clouds obtained.The result show that the algorithm structure can reliably obtain the planar and single cylinder object on it, and provide accurate data for robot grabber.