Three-dimensional point cloud semantic segmentation method based on segmentation network and adversarial network

Abstract

The invention discloses a three-dimensional point cloud semantic segmentation method based on a segmentation network and an adversarial network. The method comprises the following steps: (1) constructing the segmentation network; (2) constructing a gradient estimation module; (3) constructing an adversarial network; (4) constructing a three-dimensional point cloud semantic segmentation network based on a segmentation network and an adversarial network; (5) initializing a segmentation network and an adversarial network; (6) generating a training data set and a training label set; (7) training asegmentation network and an adversarial network; and (8) segmenting the three-dimensional point cloud data. High-dimensional features are extracted through the adversarial network, point cloud segmentation is conducted through the segmentation network, and the method has the advantages of being short in point cloud segmentation time in the test stage and high in segmentation result precision.

Description

technical field [0001] The invention belongs to the technical field of image processing, and further relates to a three-dimensional point cloud semantic segmentation method based on a segmentation network and an adversarial network in the technical field of image classification. The invention can be used to segment obstacles from the point cloud data collected by an airborne depth camera of an indoor robot, and can also be used to classify and identify urban remote sensing three-dimensional point cloud data acquired by satellites. Background technique [0002] Point cloud refers to the collection of point data on the appearance surface of the product obtained by measuring instruments. Point cloud semantic segmentation refers to assigning semantic labels to each point in the point cloud, which is a common method for spatial perception and analysis using 3D point cloud data. In the field of indoor robot obstacle avoidance, 3D point cloud data is an important carrier for recor...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to address the deficiencies of the above-mentioned prior art, and propose a 3D point cloud semantic segmentation method based on segmentation network and confrontation network, which is used to solve the existing 3D point cloud semantic segmentation method that does not take into account the segmentation results and The high-dimensional inconsistency between the real labels leads to unreasonable points in the segmentation results, and the post-processing process increases the time to segment the point cloud, resulting in a technical problem that takes a long time to segment the point cloud during the test phase in the actual application process.

Images