A three-dimensional target recognition method and system

Abstract

Embodiments of the present invention provide a three-dimensional target recognition method and system, including: for the pixel points in the initial reflected laser image, if it is determined that there is a first spatial intersection between the infrared camera projected ray corresponding to the pixel point and the target sample, then according to the first A spatial intersection point and the center point of the speckle projector to obtain the projector ray; based on the first spatial intersection point, the second spatial intersection point, the projector ray, the virtual plane and the reference speckle map, obtain the intensity value of the pixel point; according to the pixel point The intensity value of the object is obtained to obtain the final object speckle pattern, and according to the final object speckle pattern, a synthetic data set is obtained; the neural network is used to identify the object to be recognized, and the recognition result is obtained. In the embodiment of the present invention, when the training data is relatively small, by generating a synthetic data set, the requirement of a large-scale data volume of the deep learning network can be met, thereby identifying the target through the neural network.

Description

technical field [0001] The invention relates to the technical field of computer vision, in particular to a three-dimensional object recognition method and system. Background technique [0002] With the continuous development of depth perception technology and artificial intelligence technology, the monocular structured light depth imaging method based on deep learning network has a strong competitive performance by virtue of its strong learning ability and generalization ability. [0003] For the monocular structured light imaging system, research and practice have proved that the performance of the deep learning network depends on the quality and scale of the training data. Therefore, the monocular structured light depth imaging method based on the deep learning network requires large-scale and high-quality data. Each set of data required for deep learning network training includes: object speckle image, reference speckle image, and parallax ground truth (Ground Truth, GT) ...

AI Technical Summary

Problems solved by technology

[0009] Embodiments of the present invention provide a three-dimensional target recognition method and system, which are used to solve the defect that the neural network cannot be used to identify the target when the training data is small in the prior art, realize the generation of a synthetic data set, and use the synthetic data set to analyze the neural network. Perform training to identify the object to be recognized through the trained neural network

Images