Terrain classification based on variable-scale three-dimensional grid map

According to the distribution characteristics of lidar collection points, dense in the vicinity and sparse in the distance, a terrain classification method based on variable-scale three-dimensional grid map is proposed to classify an unknown terrain into four categories, which includes roads, lawns, buildings and trees. First, we establish a variable-scale three-dimensional grid map. Then the algorithm uses the point cloud feature extraction methods to extract the features of voxels. A robust outlier detection algorithm is proposed to estimate reliable local saliency features. Finally, we employ the classifiers based on TWSVM to classify the voxels into four categories. Experimental results show that our algorithm can reduce the number of voxels while ensuring accuracy, reduce the noise and have good classification results on real data sets.