Graph neural network container text recognition method based on attention mechanism

Abstract

The invention provides a graph neural network container text recognition method based on an attention mechanism. The method comprises the following steps of: S1, obtaining an image including an original container scene; S2, preprocessing the image of the original container scene to obtain a text part image in the image; S3, performing feature extraction on the text part image in the image; S4, sending extracted features to a pre-trained GTC recognition network, and recognizing text information; and S5, outputting the text information. According to the method, the processed image is input into a designed iteration correction network, the image is iteratively corrected through the same correction network, the recognition result is obtained through the GTC recognition network, and an advanced recognition effect in the world is achieved. Compared with the speed of an attention-based method, the speed of the method of the invention is greatly increased.

Description

technical field [0001] The present invention relates to the field of deep learning, and more specifically, relates to a method for recognizing container text in a graph neural network based on an attention mechanism. Background technique [0002] The container automatic identification system has a wide range of applications in customs logistics monitoring, port container management, and transportation industry container management. A container identification system with high identification accuracy is very necessary. [0003] The traditional character recognition (OCR optical character recognition) technology has been relatively mature after years of development, but it only recognizes scanned documents with a single background, high resolution and high contrast. When there are complex problems such as uneven lighting and blur in the text image of the container scene captured by the camera, satisfactory results cannot always be achieved, and manual data entry is also a very ...

AI Technical Summary

Problems solved by technology

The CTC-based method is faster due to the parallel decoding method, but the mechanism of the CTC loss function is not conducive to feature alignment and feature representation

Attention-based methods can get better alignment and feature representation, but due to their non-parallel decoding methods, such methods are slower

Images