Pointer Instrument Reading via U2-Net Dial and Pointer Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Manual reading of pointer-type instruments in harsh industrial environments is challenging due to environmental factors, necessitating an automatic reading solution that is reliable, efficient, and cost-effective.
Innovation Solution
An instrument recognition method using an improved U2 network for image segmentation, involving a two-layer nested U-shaped network structure with residual U-blocks for feature extraction, grouped convolutions, and noise reduction processing to obtain instrument readings from segmented images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual reading of pointer-type instruments is performed in harsh industrial environments, then human operators can directly read the instrument values, but the reliability and safety are compromised due to exposure to high pressure, high temperature, and high radiation environments
Solution Approach 1:
The patent replaces manual mechanical reading with an automated image processing system. A camera captures images of the instrument dial, and deep learning algorithms automatically recognize and extract the measurement values, eliminating human exposure to harsh environments while maintaining reading reliability
Solution Approach 2:
The patent introduces an intermediate image processing system between the instrument and the final reading output. The system uses image segmentation and deep learning models as intermediaries to accurately extract values without direct human intervention in hazardous conditions
2Measurement precision
If conventional image processing methods are used for instrument reading, then the system structure is simple, but the recognition accuracy is insufficient under complex environmental conditions
Solution Approach 1:
The patent extracts and segments the dial plate and pointer from the complex instrument image using deep learning-based image segmentation. This isolation of key elements improves recognition accuracy by focusing computational resources on relevant features while filtering out environmental noise
Solution Approach 2:
The patent employs a nested network structure (U2-Net with nested U-shaped blocks) where multiple processing levels are embedded within each other. This nested architecture enables multi-scale feature extraction and hierarchical processing, improving accuracy while managing computational complexity through structured organization
3Measurement precision
If deep learning-based image segmentation is implemented for instrument reading, then recognition accuracy and stability are significantly improved, but computational cost and resource requirements increase
Solution Approach 1:
The patent segments the deep learning model into specialized modules (image segmentation network for dial/pointer extraction, coordinate transformation module, value recognition module). This modular segmentation allows efficient processing by dedicating specific computational resources to specific tasks, reducing overall energy consumption while maintaining accuracy
Solution Approach 2:
The patent transforms the circular dial plate image into a two-dimensional matrix representation, enabling more efficient computational processing. This dimensional transformation optimizes the data structure for neural network processing, reducing computational complexity and energy requirements
4Productivity
If real-time monitoring is implemented in industrial production processes, then production efficiency and safety are improved, but the system requires high computational speed and low latency
Solution Approach 1:
The patent performs preliminary image segmentation and dial plate extraction before final value recognition. By pre-processing and isolating key elements in advance, the system reduces the computational burden during real-time reading, enabling faster processing speeds while maintaining accuracy for continuous monitoring applications
Data Source
AI summary
The present invention discloses an instrument recognition method based on an improved U2 network. The method includes: replacing common convolution of each layer with grouped convolution Grouped Conv on a basis of an RSU, and segmenting a dial plate and a pointer by using the network; performing noise reduction processing on the scale value array by using a mean filter; and determining a position of a scale value corresponding to the pointer by using a peak value, and outputting a reading according to the scale value and preset data. A pointer-type instrument is segmented by using the improved U2 network, and automatic reading of the obtained dial plate is implemented by using a conventional computer vision method. Therefore, compared with manual reading, the method has advantages of high precision, high reliability, fast reading, low costs, and the like, and working efficiency can be greatly improved.