Defect Detection in Partial Infrastructure Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting defects in infrastructure inspection images face challenges in managing unique defect information across large structures, as they often result in overlapping or overwritten data due to shared identification information, making it difficult to collectively manage defect information effectively.
Innovation Solution
An information processing apparatus that divides the inspection structure into partial images, assigns unique identification information to each defect, and includes outer shape information to facilitate accurate alignment and management of defect data, enabling efficient display and recording of defects without manual laborious adjustments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the entire structure is captured in a single high definition image, then defect detection quality is improved, but the image size becomes extremely large and defect detection work becomes laborious
Solution Approach 1:
The patent divides the large inspection image into multiple partial images (e.g., 4 quadrants) to reduce the processing load for each segment while maintaining overall defect detection capability. This segmentation allows the system to handle large structures without requiring manual processing of entire high-resolution images.
Solution Approach 2:
The patent introduces drawing information as an intermediary layer that overlays defect detection results on the original inspection images. This mediator enables automatic association of defects with their locations in the original image without requiring manual re-examination, thus improving detection efficiency while maintaining precision.
2Reliability
If defect information is managed in association with drawing information, then defect management is improved, but overlapping identification information makes collective management difficult
Solution Approach 1:
The patent assigns separate identification information to each partial image, creating distinct identifiers for defects found in different regions. This segmentation of identification prevents overlap and confusion when managing defects across the entire structure, as each defect can be uniquely tracked through its partial image identifier.
Solution Approach 2:
The patent creates a mapping relationship that copies defect information from partial images to the corresponding drawing information. This copying mechanism preserves unique identification while enabling centralized management, as the system maintains original identifiers while creating associated references in the drawing information database.
Data Source
Figure 1A~1F
Figure 2A~2D
Figure 3A~3B
AI summary
An information processing apparatus comprises an input unit that inputs one or more images belonging to a predetermined group, a detection unit that detects defect information based on an image inputted by the input unit, a setting unit that sets identification information that is unique within the predetermined group to the defect information detected by the detection unit, and an output unit that outputs a detection result including the defect information to which the identification information has been given by the setting unit.