Lidar Point Cloud Classification using Expectation Maximization Algorithm

EM algorithm is a common algorithm in data mining techniques. With the idea of using two iterations of E and M, the algorithm creates a model that can assign class labels to data points. In addition, EM not only optimizes the parameters of the model but also can predict device data during the iteration. Therefore, the paper focuses on researching and improving the EM algorithm to suit the LiDAR point cloud classification. Based on the idea of breaking point cloud and using the scheduling parameter for step E to help the algorithm converge faster with a shorter run time. The proposed algorithm is tested with measurement data set in Nghe An province, Vietnam for more than 92% accuracy and has faster runtime than the original EM algorithm.