Analog Meter Image Reading by Reference Point and Pointer Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing analog meter reading systems face challenges in accurately capturing measured values due to inconsistencies in imaging conditions, such as installation location and lighting, which can lead to errors in reading analog meters installed in large numbers.
Innovation Solution
An analog meter reading system that includes a unique information acquisition unit, an image acquisition unit, a detection unit, a rotation angle calculation unit, and a measured value conversion unit to detect reference points and pointers from images, reducing the influence of imaging conditions and enabling accurate measurement value reading.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If the entire analog meter is learned using classification model with 360-degree scale, then the system can process various pointer positions, but the influence of imaging conditions (lighting, installation location, capture angle) significantly degrades measurement precision
Solution Approach 1:
The patent divides the analog meter into two separate detection targets: the scale (with reference point) and the pointer. Instead of learning the entire meter including pointer positions, the system only learns to detect the scale's reference point and the pointer's presence and orientation. This segmentation isolates the detection task from imaging condition variations, as the scale and pointer are intrinsic features that maintain their relative geometric relationships regardless of lighting or capture angle.
Solution Approach 2:
The patent extracts only the essential geometric features (scale reference point and pointer) from the analog meter image, removing the need to learn the entire meter including pointer positions. By taking out only the necessary elements for measurement, the system avoids the negative influence of varying imaging conditions on the classification model while maintaining the ability to determine pointer position relative to the scale.
2Measurement precision
If operator visually checks analog meters one by one, then measurement precision can be maintained, but the productivity decreases significantly when large number of meters are installed
Solution Approach 1:
The patent replaces the mechanical/visual inspection process with an automated image processing system. The detection unit automatically identifies the scale reference point and pointer orientation from captured images, eliminating the need for manual visual checking. This substitution maintains measurement precision through consistent geometric analysis while dramatically improving productivity by enabling batch processing of multiple meters without operator fatigue or time constraints.
3Measurement precision
If classification model is created in advance with fixed imaging conditions, then measurement precision can be achieved under those conditions, but the system cannot adapt when imaging conditions change (different installation locations, lighting, capture angles)
Solution Approach 1:
The patent changes the detection parameters from learning fixed pointer positions under specific imaging conditions to detecting geometric relationships (scale reference point and pointer orientation) that are invariant to imaging conditions. By focusing on relative geometric parameters rather than absolute positions, the system maintains measurement precision across varying lighting, installation locations, and capture angles without requiring retraining or adjustment of the detection model.
Data Source
AI summary
An analog meter reading system is applied to an analog meter provided with a scale and a pointer and reads a measured value of the analog meter. The analog meter reading system includes: a unique information acquisition unit that acquires unique information of the analog meter; an image acquisition unit that acquires an image of the analog meter for reading the measured value; a detection unit that detects a reference point on the scale and a pointer from the image; a rotation angle calculation unit that calculates a rotation angle of the pointer until a state in which the pointer points to the reference point changes to a state in which the pointer points to the measuring point based on the reference point and the pointer that are detected; and a measured value conversion unit that converts the rotation angle into the measured value using the unique information.


