Wet Blue Skin Defect Segmentation for Fast Area Measurement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Manual estimation of defect areas in wet blue skin surfaces is subjective and inefficient, leading to inconsistencies and errors in quality grading.
Innovation Solution
Utilizing a Deeplab V3+ model for semantic segmentation to automatically detect and calculate defect areas in wet blue skin images, employing a MobileNetV2 network for feature extraction and dilated convolutions to enhance accuracy and real-time performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual estimation method is used to determine defect area, then quality inspectors can determine the grade of wet blue skin, but the method is subjective and inefficient with great differences in estimation by different inspectors
Solution Approach 1:
The patent replaces the manual mechanical estimation process with an automated deep learning-based image processing system. The Deeplab V3+ model automatically segments defect regions in images, and the system calculates defect areas through coordinate extraction and geometric computation, eliminating subjective human estimation and achieving consistent, precise measurements across different users.
2Reliability
If manual estimation method is used, then quality grading can be performed, but it is time-consuming and inspectors are prone to misjudgment after long work periods
Solution Approach 1:
The system enables automated self-service quality grading by capturing images of wet blue skin, automatically detecting and segmenting defects using the trained Deeplab V3+ model, calculating defect areas through coordinate processing, and determining quality grades without human intervention. This eliminates time loss and reliability issues associated with manual inspection while maintaining continuous operational capability.
3Measurement precision
If Deeplab V3+ model with MobileNetV2 and dilated convolutions is used, then defect detection accuracy and real-time performance are improved, but model complexity increases
Solution Approach 1:
The patent segments the model architecture into distinct functional components: MobileNetV2 for efficient feature extraction, dilated convolution layers for multi-scale context capture, and Deeplab V3+ for semantic segmentation. This modular segmentation allows each component to specialize in specific tasks, improving overall accuracy while enabling targeted optimization and reducing the complexity burden through structured organization.
Data Source
AI summary
Provided is an automatic calculation method for a surface defect area of a wet blue skin based on a Deeplab V3+ model, aiming at solving the problem that the automatic calculation of the defect area is difficult in a tannery. The method includes image data acquisition and preprocessing of a wet blue skin, training and verification of a wet blue skin defect segmentation model based on a Deeplab V3+ model, and automatic calculation of the defect area in a segmented image. The method can accurately segment common defects such as knife holes, brands and the like on the surface of a large-area wet blue skin, with a high speed of detecting a wet blue skin within 1s.