Terminal Inspection System Using Anchor Image Semantic Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing systems for defect detection in manufactured products face challenges in accurately identifying defects due to insufficient training data and poor training performance, leading to inaccurate inspection results.
Innovation Solution
A terminal inspection system that uses a vision device to generate digital images and a terminal inspection module performing semantic segmentation by comparing input images to anchor images, highlighting differences to identify potential defects, and utilizing an artificial neural network architecture for improved defect detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional image processing systems are used for defect detection, then the system structure is simple, but the accuracy of defect identification is poor due to insufficient training data
Solution Approach 1:
The patent uses anchor images (reference images of good parts) to train the inspection system. Instead of requiring large datasets of actual defect images, the system copies the characteristics of good parts and compares input images against these anchors to detect deviations, thereby achieving high accuracy with simpler training requirements
Solution Approach 2:
The patent replaces traditional machine learning approaches that require extensive training data with a comparison-based inspection method. The system substitutes complex neural network training mechanisms with a more straightforward anchor image comparison approach, reducing the need for large datasets while maintaining inspection accuracy
2Measurement precision
If extensive training data is collected to improve system accuracy, then defect detection performance improves, but the time required for training increases
Solution Approach 1:
The patent performs preliminary action by creating anchor images of good parts before actual inspection. These anchor images are prepared in advance and stored as reference standards, eliminating the need for time-consuming training processes during production. The system can immediately compare input images against the pre-prepared anchors
Solution Approach 2:
Instead of training the system with numerous actual defect images, the patent copies the characteristics of good parts into anchor images. This approach reduces training time significantly while maintaining high detection accuracy, as the system learns from the positive examples (good parts) rather than requiring extensive negative examples (defect images)
3Productivity
If conventional inspection methods are used, then the inspection process is simple, but operator training time and processing time are excessive
Solution Approach 1:
The patent implements self-service by enabling the inspection system to automatically compare input images against anchor images and identify defects without requiring extensive operator intervention or training. The system performs autonomous inspection operations, reducing both operator training time and processing time while maintaining high inspection efficiency
Data Source
AI summary
A terminal inspection system for a crimp machine includes a vision device configured to image a terminal being inspected and generate a digital image of the terminal. The terminal inspection system includes a terminal inspection module communicatively coupled to the vision device to receive the digital image of the terminal as an input image. The terminal inspection module has an anchor image. The terminal inspection module compares the input image to the anchor image and performs semantic segmentation between the input image and the anchor image to generate an output image. The output image shows differences between the input image and the anchor image to identify any potential defects.