Fragment Image Analysis for Crack Detection Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing work test technologies face challenges in accurately detecting cracks in tube glass due to the difficulty in training learning models with wide-range crack images, leading to excessive detection of cracks and other defects, as the images may include varying lengths and additional defects, reducing the accuracy of defect classification.
Innovation Solution
The implementation of a test system that uses a deep learning model to analyze fragment images extracted from the test image, applying filtering, binarization, and morphology processing to enhance crack detection accuracy, and an entire judgment unit to connect crack fragment images and assess the crack length against specified conditions, reducing excessive detection by inputting fragment images rather than entire defect images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the entire defect image is input to the learning model, then the defect classification can be performed, but the accuracy degrades due to wide-range images containing areas where the crack is not captured and other defects
Solution Approach 1:
The patent divides the entire defect image into multiple fragment images, each containing a specific defect. The learning model processes each fragment image separately to classify the defect type. This segmentation approach ensures that each input image focuses on a specific defect rather than a wide-range view containing multiple defects or large areas where cracks are not captured, thereby improving classification accuracy.
2Length of moving object
If the learning model is trained with wide-range crack images, then the entire crack can be captured, but the model accuracy degrades due to various crack lengths and additional defects in the image
Solution Approach 1:
The patent segments wide-range crack images into multiple fragment images, each containing a specific crack or defect. By processing fragment images individually, the learning model achieves higher accuracy for each classification while still covering the entire crack length through the combination of multiple fragment classifications. This resolves the contradiction between capturing entire cracks and maintaining model accuracy.
3Measurement precision
If fragment images are input to the learning model, then the judgment accuracy is improved, but the number of processing images increases
Solution Approach 1:
The patent segments the test image into multiple fragment images for processing. While this increases the number of images to process, it significantly improves judgment accuracy by ensuring each fragment contains a specific defect rather than wide-range views. The system manages this increased processing load through automated fragment extraction and batch processing, achieving better accuracy despite the additional computational steps.
Data Source
AI summary
Excessive detections of a defect(s) of a specified defect type is reduced. A test apparatus: inputs each of a plurality of fragment images, which are extracted from a test image of a work, into a learning model which receives an image(s) as input and outputs a type(s), and thereby judges the type with respect to each of the plurality of fragment images. The test apparatus judges whether or not a defect of a specified defect type is captured in the test image, on the basis of whether the judged type with respect to each of the plurality of fragment images is the specified defect type or not.