Simulation data set generation method and device for multi-type industrial part stacking scene

Abstract

The invention discloses a simulation data set generation method for a multi-type industrial part stacking scene, and the method comprises the steps: building a material frame of a predetermined size and three-dimensional models of different types of industrial parts, and generating a multi-type multi-instance parameterized part model library; sampling a preset number of part models from a model library; carrying out free falling motion simulation and collision simulation on the sampled part model through a dynamics simulation engine to generate a stacking scene of multi-type and multi-instance parts, and automatically labeling and storing a type label and a pose label of each object in the stacking scene; respectively generating and recording a depth image, an RGB image, a segmentation image and a complete mask image of a single object under a perspective projection view angle and an orthogonal projection view angle for each stacked scene; and repeating the steps to generate a simulation data set of the stacking scene of the various industrial parts. According to the method and device, part object models of different parameters belonging to the same parameterized template can be quickly and conveniently generated.

Description

technical field [0001] The invention relates to the field of deep learning technology, in particular to a method and a device for designing a simulation data set generation method for stacking scenes of various industrial parts. Background technique [0002] In recent years, deep learning neural network technology for industrial parts understanding has been rapidly researched and applied, such as semantic recognition, individual segmentation, pose estimation, and robotic arm grasping. Training neural networks requires a large amount of training data, but the current method of manually labeling training data sets is too cumbersome and difficult, and is prone to errors, making it difficult to effectively generate large batches of data. Therefore, a simulation data generation method with automatic labeling capability is necessary and urgent, which can greatly promote the further development of learning technology in the industrial field. [0003] The stacking scene of multiple...

AI Technical Summary

Problems solved by technology

In the stacking scene of multiple types of parts, the parts are seriously occluded from each other,

Images