Determination-Area Decision for Component Posture Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing methods for determining the position and inclination of components, such as resin-molded parts, are prone to erroneous determinations due to noise like sink marks and color differences on the surface, which affect pattern matching accuracy.
Innovation Solution
A determination-area decision method that compares feature amounts of surface shapes between different areas of a component's surfaces, identifying areas with significant differences as determination areas to reduce erroneous matching by focusing on areas with distinct edge patterns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If pattern matching is performed using the entire surface of the component, then comprehensive position and inclination identification is achieved, but erroneous determination occurs due to noise such as sink marks and color differences
Solution Approach 1:
The component surface is divided into multiple determination areas based on feature amount distribution. By segmenting the surface into regions with distinct edge patterns versus regions with noise characteristics, the system performs pattern matching only on reliable segments, thereby maintaining measurement precision while eliminating erroneous determination caused by noise areas.
Solution Approach 2:
Different regions of the component surface are assigned different qualities based on their feature amount characteristics. Areas with significant feature amount differences from the average are identified as determination areas with high reliability, while areas with minimal differences are excluded. This local quality differentiation ensures that pattern matching uses only high-quality data regions.
2Loss of information
If the entire surface is used for pattern matching, then complete component characterization is obtained, but noise affects the matching score and causes erroneous determination
Solution Approach 1:
Before performing pattern matching, the system pre-processes the component surface by calculating feature amounts across the entire surface and identifying determination areas with significant feature amount differences. This preliminary action separates useful information regions from noise-affected regions, allowing subsequent pattern matching to focus exclusively on clean data areas and thereby prevent noise from affecting the matching score.
Solution Approach 2:
The noise characteristics (sink marks, color differences) that normally degrade pattern matching are actually used as identification criteria. By calculating feature amounts and identifying areas with significant deviations from the average, the system converts noise-affected areas into identifiable determination areas or excludes them systematically, transforming the harmful noise into a structured selection criterion for reliable matching regions.
Data Source
AI summary
In the first composite step, a plurality of images is superimposed. In the second composite step, a plurality of images of a second surface of the plurality of components is superimposed. In a first detection step, a feature amount of the first surface is detected using the plurality of images superimposed in the first composite step. In a second detection step, a feature amount of the second surface corresponding to the plurality of areas of the first surface is detected using the plurality of images superimposed in the second composite step. A difference in the feature amounts between each area of the first surface and each area of the second surface corresponding to each area of the first surface is calculated. In the determination-area decision step, an area where the difference in the feature amounts calculated in the calculation step is greater than a predetermined value is decided.