A CNN-based method for identifying the number of spliced ​​surfaces contained in non-calibrated surfaces

Abstract

The invention discloses a method for identifying the number of spliced curved surfaces in an uncalibrated curved surface based on a CNN (convolutional neural network). The method includes the steps: reducing dimensions of a five-dimensional point cloud group of a target curved surface to be identified to two dimensions to obtain a two-dimensional point cloud group of the target curved surface to be identified by a PCA (principal component analysis) method, and taking the two-dimensional point cloud group of the target curved surface to be identified as one group of target curved surface data to be identified; inputting the target curved surface data to be identified into a successfully trained CNN spliced curved surface number identifier, and outputting the number of the spliced curved surfaces in the target curved surface to be identified. According to the method, the number of the spliced curved surfaces in the target curved surface to be identified is outputted according to the successfully trained CNN spliced curved surface number identifier, and the method is rapid in identification and high in identification accuracy. Besides, the dimensions of the five-dimensional point cloud group are reduced by the PCA method, and calculated amount is reduced to a large extent, so that identification is accelerated.

Description

technical field [0001] The invention relates to the field of reverse engineering, in particular to a CNN-based method for identifying the number of spliced ​​curved surfaces included in non-calibrated curved surfaces. Background technique [0002] Reverse engineering is based on the existing product model, using digital measurement equipment such as laser scanners to obtain physical data such as point clouds, and then segmenting and fitting these physical data to build a complete CAD model. [0003] Because point cloud has the characteristics of convenient storage and flexible calculation, it is an important form of metadata in computer graphics, and point cloud reverse reconstruction is an important technology in point cloud computing. In recent years, with the continuous improvement of digitization precision and the complexity of scanning object surfaces, it is very time-consuming to segment massive point clouds acquired by equipment such as 3D laser scanners or CT scanner...

AI Technical Summary

Problems solved by technology

In recent years, with the continuous improvement of digitization accuracy and the complexity of scanning object surfaces, it is very time-consuming to segment massive point clouds acquired by equipment such as 3D laser scanners or CT scanners.

Images