A kernel-density-estimation-based outlier detection for airborne LiDAR point clouds

An outlier detection method is proposed based on the kernel density estimation for removing the outliers in airborne LiDAR point clouds. The point cloud is divided into many blocks. Then, in each block, the kernel probability density of the height values of all points is estimated. Two elevation thresholds, one for low outliers and one for high outliers, are selected based on the values of the probability density and the values of elevation. The computation is simplified in complexity for the method doses not focus on the calculation of individual points. Two datasets were utilized to test our method. This method combines distance-based method with density-based method. Experiments showed that our proposed method had good performance.