Digital instrument reading recognition method, system and readable storage medium

By using a lightweight deep residual neural network and adaptive local contrast enhancement processing, the problems of uneven illumination and confusion of highly similar characters in mechanical water meter and electricity meter reading recognition systems are solved, achieving high-accuracy reading recognition and reducing equipment resource requirements.

CN122336764APending Publication Date: 2026-07-03SHANGHAI FULLHAN MICROELECTRONICS

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
SHANGHAI FULLHAN MICROELECTRONICS
Filing Date
2026-03-31
Publication Date
2026-07-03

Smart Images

  • Figure CN122336764A_ABST
    Figure CN122336764A_ABST
Patent Text Reader

Abstract

This invention provides a method, system, and readable storage medium for recognizing digital instrument readings. The method includes acquiring an image of a digital region on a digital instrument dial; segmenting the digital region image according to the digital position to obtain a single-character region image at the corresponding position; for each single-character region image, performing the following operations: adjusting the single-character region image to a standardized single-character image of a preset size; performing adaptive local contrast enhancement processing on the standardized single-character image to obtain an enhanced single-character image; and recognizing each enhanced single-character image based on a lightweight deep residual neural network, outputting the recognition result. This invention effectively solves the problem of loss of local details of digital characters caused by reflections, shadows, or low illumination, and has high reading recognition accuracy. The model computational complexity and parameter count are relatively small, and the requirements for device resources are low, allowing for smooth deployment in microcontroller-level or low-power vision SOC chips.
Need to check novelty before this filing date? Find Prior Art