Reverse modelling method of transmission tower based on intelligent identification of key points of point cloud projection

At present, neither image-based oblique photographic modelling nor laser point cloud modelling can generate high-precision solid models of transmission towers. In this paper, a fast reverse modelling method based on deep learning for the identification of key points of the transmission tower point cloud plane projection and high accuracy is proposed. This method first transforms the 3D point cloud data into two-dimensional image data through orthographic projection, then uses CNN and LSTM two deep learning networks to predict the number of variable key points in the image and finally calculates the corresponding coordinates in 3D space through the key points in the twodimensional image. Through this method, the size and position of transmission tower components are obtained, which can realise fast and high-precision reverse modelling of transmission tower. © 2023 Inderscience Enterprises Ltd.