Application of Structured Light 3D Reconstruction Technology in Industrial Automation Scenarios in the Context of Digital Transformation

Abstract

In this work, a dynamic structured light 3D reconstruction method adapted to random phase shift step size is proposed for the problem of dynamic workpiece 3D reconstruction in industrial manufacturing scenes. The method firstly utilizes a structured light sinusoidal grating pattern to match the relative phase shift of the workpieces moving along the perpendicular direction of the stripes on the conveyor belt. Secondly, in order to solve the problem of non-uniform phase shift step length due to external vibration, electromagnetic radiation and other disturbing factors of the conveyor belt and other mechanical devices, the method proposes a dynamic structured light 3D reconstruction model based on RPSNet, which is based on the CycleGAN network model, and uses the AIR2U-net model proposed in this paper as the generator, and the multilayer convolutional neural network CNN as the discriminator to realize the grating map. discriminator to realize the conversion of raster map to depth map. Finally, for the problem that the network model needs a large amount of data for training and it is difficult to collect data in the actual industrial scene, this paper uses the Thing10k dataset, and the dataset made by Blender simulation software for model training. Finally, a higher quality 3D reconstruction of the workpiece is realized.