FCN Crack Detection in Nuclear Video via Patch Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automated crack detection algorithms are inefficient for real-time analysis of high-resolution videos from nuclear power plant inspections, often resulting in long processing times and high false positive rates due to small crack sizes and complex surface textures.
Innovation Solution
A fully convolutional network (FCN) architecture with a parametric data fusion scheme is used to analyze video frames and generate crack score maps, allowing for rapid detection of cracks in full-HD videos by leveraging spatiotemporal coherence and requiring less processing time than existing methods.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing automated crack detection algorithms (LBP-SVM, NB-CNN) are used to analyze full-HD video frames, then crack detection accuracy is improved, but processing time increases to 12.58-17.15 seconds per frame
Solution Approach 1:
The video frame processing is segmented into multiple overlapping patches that are processed independently and then merged. This allows parallel processing of different regions, reducing overall processing time while maintaining detection accuracy through the overlapping strategy that ensures comprehensive crack detection.
Solution Approach 2:
Instead of processing the entire full-HD frame at once, the algorithm processes only relevant patches containing potential crack features. The overlapping patch strategy applies partial action to critical regions while avoiding redundant processing of entire frames, thus reducing processing time without sacrificing detection accuracy.
2Productivity
If conventional crack detection algorithms (edge detection, thresholding, morphological operations) are used, then processing speed is improved, but false positive rate increases due to small crack sizes and low contrast on metallic surfaces
Solution Approach 1:
The algorithm transforms the detection parameters by converting raw pixel values into LBP texture descriptors and then into SVM feature vectors. This parameter transformation enables the system to distinguish subtle crack patterns from surface textures by operating in a transformed feature space rather than raw pixel space, reducing false positives while maintaining processing speed.
Solution Approach 2:
The patent replaces conventional mechanical processing operations (edge detection, thresholding, morphological operations) with a machine learning-based approach (LBP-SVM). This substitution allows the system to automatically learn and adapt to the specific characteristics of cracks on metallic surfaces, significantly reducing false positives caused by surface textures like scratches, welds, and grind marks.
3Reliability
If manual inspection methods are used to review optical images or video of components, then detection reliability is improved, but inspection time and human error increase when reviewing large quantities of data
Solution Approach 1:
The system performs self-service by automatically detecting and classifying cracks without requiring manual review. The LBP-SVM algorithm autonomously processes video frames, identifies crack patterns, and generates detection results, eliminating the need for human inspectors to review large quantities of data while maintaining high detection reliability and significantly reducing inspection time.
Data Source
AI summary
Systems and methods for detecting cracks in a surface by analyzing a video, including a full-HD video, of the surface. The video contains successive frames, wherein individual frames of overlapping consecutive pairs of the successive frames have overlapping areas and a crack that appears in a first individual frame of a consecutive pair of the successive frames also appears in at least a second individual frame of the consecutive pair. A fully convolutional network (FCN) architecture implemented on a processing device is then used to analyze at least some of the individual frames of the video to generate crack score maps for the individual frames, and a parametric data fusion scheme implemented on a processing device is used to fuse crack scores of the crack score maps of the individual frames to identify cracks in the individual frames.