Classification of Pavement Disease 3D Point Cloud Images Based on Deep Learning Network

In order to solve the problem that the linear array image of the two-dimensional road surface is easily affected by factors such as illumination changes, shadows, this thesis carried out the research on the design of the intelligent perception system of the three-dimensional information of the highway road surface. This paper first analyzes the design requirements of the intelligent perception system for 3D information on highway pavement, studies the cooperative working mechanism of various high-precision sensors, and proposes a hardware system architecture scheme for 3D information perception of pavement diseases based on multiple sensors. Then, a method of “blocking” classification of road 3D point cloud images based on deep learning network model is proposed. Based on the construction of a small block road point cloud data set, an experimental comparative study is carried out on Pointnet and Pointnet++ networks. The experimental results show that the recognition rate of pavement diseases is better than 86.7% and 90.0%.