Image-based text recognition method, system, device and storage medium

Abstract

The present invention provides an image-based text recognition method, system, device, and storage medium. The method includes: obtaining at least one frame to be detected containing a partial image and the corresponding frame to be detected from the image through the first deep learning model according to the first training set. The first annotation of the frame, the first training set includes a plurality of first-type sub-training sets, and each first-type sub-training set includes the first annotation of an image of a character group; separately use the corresponding to each frame to be detected A sub-training set of the second type obtains the character label in the frame to be detected and the second position information of the character in the image through the second deep learning model, and each sub-training set of the second type corresponds to a sub-training set of the first type A second label of each character in the character group; arranging the characters according to the second position information to obtain a character string. The invention can reduce model complexity, calculation amount and storage space, and improve the overall speed and accuracy of model detection.

Description

technical field [0001] The invention relates to the field of gate security inspection, in particular to an image-based text recognition method, system, device and storage medium. Background technique [0002] The port is the distribution center of foreign trade import and export goods, an important link in the international logistics supply chain and the hub of logistics channels. Today's smart port construction in my country has pushed the port field into a critical period of digital transformation. Based on manual and traditional optical character recognition method, it is difficult to meet the needs of intelligent and efficient port management at present, and the intelligent recognition scheme based on deep learning can greatly reduce the labor intensity. Advantages, especially less susceptible to complex working conditions such as light, rain and snow, wind and sand, and camera lens defacement that reduce image clarity. [0003] Traditional image recognition based on deep...

AI Technical Summary

Problems solved by technology

[0002] The port is the distribution center of foreign trade import and export goods, an important link in the international logistics supply chain and the hub of logistics channels. Today's smart port construction in my country has pushed the port field into a critical period of digital transformation. Based on manual and traditional optical character recognition method, it is difficult to meet the needs of intelligent and efficient port management at present, and the intelligent recognition scheme based on deep learning can greatly reduce the labor intensity. Advantages, especially less susceptible to complex working conditions such as light, rain and snow, wind and sand, camera lens contamination, etc. that reduce image clarity

[0003] Traditional image recognition based on deep learning needs to perform convolution processing on the whole image, which has a huge amount of calculation and high cost. Because it is a full-image convolution, the output labels must include Chinese characters, English alphabetic characters, numeric characters, and punctuation marks. And so on, a variety of annotations, greatly increasing the amount of calculations, and also increasing the possibility of errors in annotations

Moreover, when some vehicles pass through the gate at high speed, the quality of the photos taken is not high, and if the whole picture is still processed, the system will not be able to process the data in time, which will reduce the accuracy of recognition.

Images