Abstract:Due to the narrow space of underground Wells and complex environment, the collected site cloud data has irregular and uneven distribution in the space and large amount of data, which makes it difficult to achieve multi-objective and efficient semantic segmentation. In this paper, a semantic segmentation model of site cloud based on PointNet++ is proposed. Firstly, the data of underground well point cloud is collected and the data set required for model training is added to the semantic label. Secondly, to improve the performance of the high cloud segmentation task, a deep learning-based PointNet++ network model is introduced and a strategy of Multi-scale grouping (MSG) and Random Input Dropout (DP) is utilized. The semantic segmentation of ground, roof, ladder, wall, cable and support etc. is realized. Finally, accuracy, recall rate, crossover ratio and F1 score are used as evaluation indexes for segmentation. The results show that compared with PointNet network model, the evaluation indexes of this model have shown significant improvement, and the evaluation indexes of surface, well top, well wall and cable are all over 80%, indicating good performance of target segmentation. It is conducive to multi-objective fast and accurate segmentation of underground cable well scene, and lays a foundation for fine management of underground engineering.