Research On Indoor Obstacle Measurement And Three-dimensional Reconstruction System Of Mobile Robot

With the continuous development of science and technology and the improvement of people's lives,Three-dimensional reconstruction technology attracts more and more attention in the modern society.With the application of sensor device to acquire the depth information of the scene,we can realize the target scene reconstruction.It has a wide application prospects in the field of the cultural relic protection,the virtual military drills,the three-dimensional video entertainment,medical rehabilitation.During the early stage,with the corresponding algorithm,the technology of three-dimensional reconstruction obtains depth information for three-dimensional reconstruction of objects with laser scanning device as a sensor.However,this early-stage technology has several disadvantages,such as the high expense of equipment,the large amount of data,high degree of difficulty in development and long processing time.In order to solve the above problems,this topic collects the three-dimensional reconstruction data by the low price Kinect,builds three-dimensional reconstruction system based on Kinect data,and rebuilds the actual scene and so on.Firstly,in terms of access to the original data,The Kinect depth sensor is used to measure the reconstruction of object or scene.OpenNI interface is used to drive the Kinect to collect the original data.Two-dimensional image information can be converted into three-dimensional point cloud for reconstruction by use of the relationship between the image coordinates,camera coordinates and world coordinates.Secondly,the data obtained with Kinect is in different coordinates and there are rotation and translation among the data.It is necessary to match the feature points in the image to obtain accurate matching points,which provides better initial condition for the next point cloud fusion.By analyzing and improving the commonly used matching algorithms,we adopt a method based on SURF,bidirectional FLANN algorithm and improved RANSAC algorithm to extract image features.Comparative experiments are carried out.After successful obtaintion of the matching point,the point cloud data can be fused.The traditional ICP algorithm is improved from two aspects of feature matching and initial condition.Finally,experiments are carried out to verify the proposed reconstruction method.The results indicate that the proposed method has the ability of high-speed processing,reduces the reconstruction time and can achieve better performance in reconstruction.