Defect Detection Using Dimensionality Reduction Lighting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automated systems for inspecting parts for visible defects are limited by the combinations of lighting patterns they can produce and require a large number of images, making them slow and expensive in production line settings.
Innovation Solution
A method and apparatus that use a hollow structure with distributed light sources and a camera, where the lighting combinations and intensities are determined through a setup phase involving dimensionality reduction and image processing to efficiently detect defects, allowing for quick and accurate inspection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a large number of images are taken to ensure rigorous inspection, then detection accuracy is improved, but inspection speed deteriorates
Solution Approach 1:
The system performs a setup phase before actual inspection to determine the optimal subset of light source combinations. During this preliminary action, the system analyzes sample images and performs dimensionality reduction to identify which lighting patterns provide the most information for defect detection. This pre-computed knowledge allows the inspection phase to use only the necessary number of images with optimal lighting, achieving high accuracy without requiring a large number of images, thus resolving the contradiction between detection accuracy and inspection speed
Solution Approach 2:
The system extracts only the essential information needed for defect detection by performing dimensionality reduction on the N-dimensional images. It identifies and discards redundant lighting patterns that do not contribute significantly to defect detection. By taking out only the necessary components (essential light source combinations and their intensities), the system reduces the number of images required from N to n, where n < N, thereby maintaining detection accuracy while improving inspection speed
2Measurement precision
If multiple lighting patterns are used to improve defect detection, then detection accuracy is improved, but system complexity increases
Solution Approach 1:
The system extracts and identifies only the essential light source combinations and their corresponding intensities that provide the most valuable information for defect detection. Through dimensionality reduction and analysis of sample images, it determines a reduced set of n lighting patterns from the original N possible patterns. This extraction of essential elements maintains high detection accuracy while reducing the complexity of the lighting control system and image processing requirements
Solution Approach 2:
The system changes the parameters of the lighting system by determining optimal combinations of light sources and their intensities through the setup phase. Instead of using all possible lighting patterns, it identifies specific parameter settings (which light sources to activate and at what intensity) that maximize defect detection capability. This parameter optimization allows the system to achieve high detection accuracy with a manageable number of lighting configurations, thereby reducing system complexity
3Measurement precision
If N-dimensional images are processed to maintain full information, then detection accuracy is improved, but processing time increases
Solution Approach 1:
The system extracts the essential information from N-dimensional images by performing dimensionality reduction. It identifies the minimum number n of eigenvectors (where n < N) that contain sufficient information for defect detection. By extracting only this essential subset, the system reduces the processing load and memory requirements while maintaining the ability to detect defects accurately, thus reducing processing time without sacrificing detection accuracy
Solution Approach 2:
The system applies partial action by processing only the necessary portion of the image data. Instead of fully processing all N-dimensional image information, it processes a reduced n-dimensional representation that contains the critical defect detection information. This partial processing approach is sufficient for the intended purpose (defect detection) while significantly reducing computation time and resource requirements
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The method enables fast and efficient detection of visible defects with high accuracy by determining optimal lighting conditions and reducing the number of images required, while ensuring the lighting conditions are within the capabilities of the apparatus.
Implementation Method 1
illuminating the part to be tested with the light sources in said predetermined combinations at said predetermined intensities; imaging the part to be tested synchronously with illuminating said part to be tested
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Method of detection of visible defects in physical parts comprising: - providing an apparatus (1) comprising a hollow structure (2) adapted to receive a part (50), a plurality of light sources (3) distributed on the inside of the hollow structure (2), a camera (7) for imaging the part (50), and a controller (5) for selectively illuminating at least a portion of said light sources (3) in predetermined combinations at predetermined intensities and for actuating said camera (7); said method comprising an operational phase comprising the steps of: - providing a part (50) to be tested in the interior of said enclosure; - illuminating the part (50) to be tested with the light sources (3) in said predetermined combinations at said predetermined intensities; - imaging the part (50) to be tested synchronously with illuminating said part (50) to be tested, thereby obtaining a number n of original test images (51); - processing said original test images to detect visible defects in said part to be tested. According to the invention, said predetermined combinations of light sources and said predetermined intensities are determined in a setup phase by: - imaging a number of parts (50) without visible defects and a number of parts (50) with known visible defects while illuminating each part (50) with individual light sources (3) or groups of light sources (3) at known intensity, thereby obtaining a number N of sample images corresponding to a given part; - identifying visible defects in said sample images; - combining said sample images for each part into an N-dimensional image for each part having N dimensions per pixel; - performing a dimensionality reduction of said N-dimensional images to reduce the dimensionality of said images while still permitting identification of said visible defects, thereby generating corresponding n-dimensional images; - determining said predetermined combinations of light sources and said predetermined intensities based on a result of said dimensionality reduction.