3D point clouds recognition method for substation equipment using a new attribute descriptor method

As smart grids evolve, substation point cloud data is vital for management, maintenance, and monitoring. However, traditional recognition algorithms struggle with challenges such as noise, occlusion, viewpoint changes, and uneven density. This paper proposes a novel approach for substation equipment point cloud recognition. It first establishes a local coordinate system for the point cloud based on symmetry and spatial distribution, achieving translation and rotation invariance. A new attribute descriptor is then defined, considering form characteristics and viewpoint variations, and a template database of 20 electrical devices is created. By matching descriptor of a device's point cloud with those in the template database, the correct device is identified. The proposed method is compared with two other methods, achieving a 90% identification precision with an average time of 3.2 seconds per device. The method demonstrates robustness, maintaining over 70% accuracy even with noise, occlusion, and irregular point cloud density.