A connected component labeling algorithm for sparse Lidar data segmentation

This paper proposes an extended connected-components labeling algorithm for sparse Lidar (Light detection and ranging) sensor data. It is difficult to label sparse Lidar data using the general connected-component labeling algorithm. The proposed technique first increases the density of the sparse data by performing mathematical morphological operation of dilation. Next, labeling is performed on the dilated data, and the resultant labels are mapped to the input sparse Lidar data. The proposed technique does not distort the input Lidar data. We show the application of the proposed algorithm in map building using clustering. Results show that the proposed method can label sparse Lidar data to build maps.