Research On 3D Point Cloud Map Construction Using 3D Laser Scanner

Coal has been the largest non-renewable resources in our country's energy consumption, but underground mining's mechanization and intelligence is low now. So it need to establish a remotely monitoring system to achieve visual monitoring, and to provide technical support for thin coal seam unmanned mining in the complex environment.The main content of this paper is building underground three-dimensional point cloud map, an important part of visual monitoring. To build this map, firstly using homemade three-dimensional laser scanner to obtain 360o point-cloud data; then filtering noisethrough the point-cloud's straight filtering and neighbor filtering, and reducing the number of point cloud to obtain the effective registration data by down-sampling point cloud. However, two existing problems of the aforementioned process are as follows, one is standard ICP(Iterative Closest Point) registration algorithm is low search efficiencyand easy to falling into local optimal solution, For solving these problems, we can use k-d tree accelerating ICP search efficiency, and use SAC-IA(Sample Consensus Initial Alignment) algorithm for rough registration and improving ICP registration algorithm's accuracy to avoid local optimal solution, achieving accurate matching for two pointcloud, and analyzingthe efficiency and high accuracyand robustness of improved ICP registration algorithm, based onSAC-IA. The other is point cloud registration accumulated error. The more number of point cloud registration, the greater the accumulated error, and lead to the final three-dimensional map's serious distortion.Therefore,this paper put forward a optimizedscheme which combines map relaxation and Moving Least Square algorithmto optimize the 3D map, and manage it to inquire and call date conveniently.Finally, based on the above study, this paper designs and realizes a 3D map construction system to meet coal mining's production requirements. Using this system, we can achieve key functions of point cloud's filtering, down-sampling,registration, construction and so on.Alsousing 3D map generated by the system to planningpath and cutting window measure.