Research On Key Technologies Of 3D Map Building Based On Kinect In Indoor Environment

With the rapid development of science and technology and the rapid increase of social demand,service robots are gradually entering people's lives,and changing people's way of life.Seeing the huge market potential of the service robot industry,governments,scholars and entrepreneurs of all countries have devoted great enthusiasm to the research and development of service robots.For service robots,autonomous navigation is the key to its application,and it is also the most important technology of autonomous mobile and intelligent.In order to realize localization and navigation,the robot needs to be aware of the surrounding environment,so the environmental map information plays a very important role in the autonomous navigation of the robot.At present,the demand of 3D map becomes more extensive and more accord with human life.Also with the development of technology,the research and development of 2D map to 3D map is an inevitable trend.So this paper launches the research on the key technologies of 3D map building based on Kinect.First of all,in the VS2012 development environment combined with OpenCV computer vision library,based on Zhang Zhengyou plane calibration method to complete the camera calibration,and ultimately generate point cloud data.Secondly,in order to improve the speed and accuracy of point cloud registration,it is necessary to denoise and simplify the point cloud data obtained by Kinect.In the part of point cloud denoising,the statistical filter based on statistical analysis theory is used to remove noise points and outliers;In the part of point cloud simplification,a simplified method of point cloud is proposed based on the combination of octree space division and curvature characteristics.According to the different curvature,the point cloud data can be divided into the characteristic region and the non-characteristic region.For the feature area,the minimum distance method is used to streamline,and for the non-characteristic region,the octree is used to reduce the uniformity.The experimental results show that this method not only has a good reduction rate,but also can maintain the characteristic information of the point cloud.Thirdly,the RANSAC initial registration method based on curvature feature is proposed.Firstly,the feature points are extracted according to the curvature feature information,and then the initial registration is realized according to RANSAC.On the basis of the initial registration,this paper presents a ICP algorithm based on the curvature feature in the overlap region of the two sets of points cloud.The experiment platform mainly includes two parts: personal computer and Kinect sensor.In the VS2012 development environment,combined with point cloud database to complete the point cloud simplification and registration of the experimental part.The feasibility and effectiveness of the proposed method is verified.Finally,the registration effect of interior point cloud data is demonstrated.