Laser scanning SLAM indoor three-dimensional point cloud quality evaluation method based on deep learning

Abstract

The invention discloses a laser scanning SLAM indoor three-dimensional point cloud quality evaluation method based on deep learning. The method comprises the steps that S1, acquiring high-quality point cloud through a laser scanning SLAM device; S2, performing degradation on the high-quality point cloud to obtain a simulation point cloud; S3, carrying out track measurement analysis on the simulation point cloud; S4, extracting a plane from the high-quality point cloud and the simulation point cloud, performing local consistency noise analysis and geometric rule analysis on the plane, and quantifying the quality of the point cloud; S5, segmenting the high-quality point cloud and the simulation point cloud to obtain point cloud blocks; S6, normalizing the point cloud blocks and then inputting into a Point Net + + neural network for model training, and obtaining a network model; S7, performing point cloud quality analysis on the point cloud to be evaluated through the step S4 to obtain a point cloud quality level value; and S8, predicting the to-be-evaluated point cloud through the neural network model obtained in the step S6, and judging whether the point cloud belongs to a high-quality point cloud or a quality-reduced point cloud. The invention provides a method for quantifying the quality of point cloud, and establishes a classification standard and a framework for evaluating an indoor three-dimensional point cloud model under an SLAM system.

Description

technical field [0001] The invention relates to the field of quality evaluation of three-dimensional point clouds, in particular to a deep learning-based laser scanning SLAM indoor three-dimensional point cloud quality evaluation method. Background technique [0002] In recent years, in order to meet people's needs for indoor space information, positioning and building modeling, laser scanning technology is also being applied to indoor 3D model reconstruction and drawing. Typically, indoor laser mapping usually employs a mobile mapping system. At present, the mobile mapping system generally adopts the Simultaneous Localization and Mapping (SLAM, Simultaneous Localization and Mapping) method, which can reconstruct or even restore the outline of the indoor scene in a short time. Compared with the static scanner on the ground, the mobile mapping system can move through the platform, obtain environmental data from multiple angles, obtain the point cloud of the three-dimensional...

AI Technical Summary

Problems solved by technology

For example, around the edge of a sharp object, there may be greater noise, including higher incident angles, distance from the object, and noise interference caused by building materials with poor reflectivity

[0010] (2) Secondly, the quality analysis of indoor point clouds focuses on the accuracy description, and does not form a complete framework and system

[0011] (3) Third, due to the diversity of data sources and differences in SLAM algorithms, it increases the difficulty of evaluating the quality of point clouds

It is not objective enough to estimate the accuracy of the SLAM system by only using the trajectory evaluation standard generated by the SLAM algorithm, and the point cloud map generated by the SLAM algorithm cannot be accurately evaluated.

Images