Monocular structured light depth imaging method based on convolutional neural network

Abstract

The embodiment of the invention provides a monocular structured light depth imaging method based on a convolutional neural network. The method comprises the following steps: S1, carrying out visual calibration on a monocular structured light imaging hardware system; s2, generating a synthetic training data set; s3, carrying out data augmentation and preprocessing; s4, constructing a convolutionalneural network; s5, determining a loss function; s6, setting an optimizer; s7, training a convolutional neural network; and S8, testing the convolutional neural network. The method provided by the embodiment of the invention is used for training a synthetic training data set of a convolutional neural network, specific optimization and improvement are carried out on face recognition, the imaging effect and the recognition accuracy when monocular structure light depth imaging is adopted for face recognition are improved, the network calculation amount is reduced, and the real-time performance under the imaging quality requirements for high precision and high resolution is improved.

Description

technical field [0001] The invention relates to the technical field of machine vision, in particular to a convolutional neural network-based monocular structured light depth imaging method. Background technique [0002] Depth perception imaging technology has always been an important research direction and topic in the field of machine vision. Among them, the monocular depth imaging technology based on spatially encoded structured light is one of the most mainstream directions for short-range depth imaging, and is widely used in consumer electronics, somatosensory games, security and other scenarios. [0003] The existing monocular structured light depth imaging algorithm cannot be specifically optimized for face recognition, and the accuracy of face recognition is low, which cannot meet the increasing quality requirements of current face recognition applications for depth imaging. At the same time, for the face recognition application on the embedded platform, the existing...

AI Technical Summary

Problems solved by technology

[0004] An embodiment of the present invention provides a monocular structured light depth imaging method based on a convolutional neural network, which is used to solve the poor imaging effect and low recognition accuracy when using the existing monocular structured light depth imaging method for face recognition. The problem of poor real-time performance under high-precision and high-resolution imaging quality requirements

Images