Research On LiDAR-based 3D Point Cloud Data Registration And Application

With the increasing demand of unmanned combat system in the future information war and the development of sensor,the demand of the unmanned vehicle environmental perception,augmented reality,environmental modeling and many other technical requirements are also increasing.The study on LiDAR of environmental perception,environmental modeling based on unmanned vehicle and so on is essential.The problem of how to solve the 3D point cloud registration of Li DAR in non-structural environment is the most important problem in this paper.The real unstructured environment is complex,dynamic and full of uncertainty,and it is not just a mathematical model can describe.The mobile of unmanned system is not accurate and can be easily disturbed by the surrounding environment,time and many other factors,but also full of noise.In summary,a new method is needed to solve the registration problem in non-structured environment.In this paper,the research and improvement of the registration algorithm based on the characteristics of unmanned ground systems are studied,and the construction of the LiDAR point cloud map is finally completed.The research results of this paper include:1 An improved ICP(Closest Point Iterative)algorithm based on image feature extraction is proposed.The basic idea of this algorithm is to improve the registration point set in ICP registration algorithm,which is based on a series of feature points from point cloud data and image data fusion.The advantage of this algorithm is that it can reduce the time of registration by improving the accuracy of registration and the number of points on the other hand,so as to overcome the shortcomings of the traditional ICP algorithm.Through the experiment,we can get the conclusion that the algorithm can improve the accuracy and efficiency.2.Using a based on feature extraction ICP and LOAM(lidar Odometry and mapping in real time)built map fusion algorithm of point cloud map construction method,this method through based on image feature extraction improved ICP algorithm for point cloud of registration,then using LOAM algorithm built Charcot Marie Tooth to real time on cloud point position is corrected,and the corrected results are updated,the update of the point cloud as a map construction using the point cloud of incremental mapping,to limit the number of physical point,to complete the construction of the point cloud map.Through experiments,the method can be proved that it construct a good point cloud map.