Meter reading model training method and meter reading method

Abstract

The invention relates to the technical field of image recognition, in particular to a meter reading model training method and a meter reading method.The meter reading model training method comprises the following steps that a first image with the reading number of a camera direct-reading meter is obtained, the first image comprises multiple characters with the high-low position relation, and the multiple characters comprise the whole character and the half character; marking the first image as a meter reading numerical value corresponding to the plurality of characters; inputting the first image into a convolutional neural network model for training to obtain a training result; and adjusting parameters of the convolutional neural network model according to the training result and the marked meter reading value of the first image. The method has the effects of improving the recognition accuracy of the reading of the camera direct-reading meter and simplifying the working process of the algorithm.

Description

technical field [0001] The application relates to the technical field of image recognition, in particular to a meter reading model training method and a meter reading method. Background technique [0002] Water meters, electricity meters, etc. can use camera direct reading meters. Taking the water meter as an example, the camera direct-reading meter recognizes the numbers on the water meter at the meter end, and uploads the recognized data to the platform, thereby reducing the labor cost of traditional water meters that require manual meter reading, and Improved meter reading efficiency. [0003] The OCR (Optical Character Recognition) method is used for the recognition of the readings of water meters and electricity meters in the camera direct reading remote meter reading system. First, the feature extraction is performed on the digital part of the characters, so as to locate the upper and lower parts of a single character on the water meter. The left and right boundaries...

AI Technical Summary

Problems solved by technology

[0004] The OCR recognition method in the related art cannot accurately locate and recognize water meters in complex scenes, such as uneven lighting, blurred pictures, and defaced pictures, and the accuracy of character segmentation is very high in the recognition process. , and heavily depends on the establishment of the font library, and the ORC recognition method in the related technology is not effective in recognizing half-characters on the water meter. When encountering multi-characters with high-low relationship and about to be carried, the recognition often has a large error

Images