3D point cloud semantic segmentation migration method based on meta-learning

Abstract

The invention discloses a 3D point cloud semantic segmentation and migration method based on meta-learning, and relates to the technical field of robot navigation. The method comprises the following steps: constructing a Point Net network model; selecting a training data set; for each training data set, forming a training task set by utilizing different types of data; constructing a meta-learningframework; according to the meta-learning framework, training the Point Net network model through each training task set; selecting a test task set; and inputting the test task set into the trained Point Net network model for testing until the gradient update value of the model converges. According to the method, a trained model is loaded in a new environment task, optimal similar task parametersare used, and the training efficiency of the indoor scene semantic segmentation method of the new task is high; the semantic segmentation capability of different tasks is learned through a meta-learning framework, it is ensured that learning features can be suitable for different migration environments, and the generalization performance of the model is improved.

Description

technical field [0001] The present invention relates to the technical field of robot navigation, in particular to a method for semantic segmentation and migration of 3D point clouds based on meta-learning. Background technique [0002] In the field of robot navigation technology, semantic segmentation is a fine-grained classification task, a basic problem in realizing scene understanding in computer vision tasks, and an important step in realizing the robot from rough reasoning to fine reasoning. [0003] At present, the semantic segmentation method of 3D point cloud data is developed from the semantic segmentation of images. According to the different expression forms of data, there are mainly the following two methods: [0004] Multi-view projection: project 3D point cloud data onto different planes to obtain 2D images from different perspectives, process 2D images using image semantic segmentation methods, and then fuse data information from different perspectives to obta...

AI Technical Summary

Problems solved by technology

Due to the uneven data distribution caused by the 3D point cloud data acquisition itself, the sparsity of the data will affect the accuracy of feature extraction. At the same time, the 3D convolution operation has a large amount of calculation and low data processing efficiency, so this method is difficult to apply

[0006] The above two methods cannot directly use point cloud data, and need to convert point cloud data. This process has information loss and low efficiency; in addition, the amount of labeling data required is large, especially for semantic segmentation tasks, pixel-level data is required Labeling, heavy workload and high cost

Images