Dual-Model Image Inspection for Inkjet Defect Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image inspection systems for inkjet recording devices face challenges in detecting defects in recorded images due to variations in image types, such as text, geometric shapes, and photographs, which can lead to decreased precision in defect detection.
Innovation Solution
An image inspection system that utilizes two trained models generated by machine learning. The first model analyzes recorded images to detect defects, while the second model evaluates recording information related to the device and medium to provide additional context for defect detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If only image data is used as input for machine learning-based defect detection, then the system is simple and easy to operate, but the detection precision deteriorates when detecting defects in image types not learned during training
Solution Approach 1:
The input data for defect detection is segmented into two distinct types: image data and recording information data. The first trained model processes image data while the second trained model processes recording information, allowing each model to specialize in its data type and improve overall detection precision without requiring one model to handle all data complexity
Solution Approach 2:
The system transitions from using only two-dimensional image data to incorporating a third dimension of data by adding recording information (one-dimensional tabular data about recording conditions). This multi-dimensional approach enables the system to detect defects more accurately across various image types by considering both visual patterns and contextual recording parameters
2Measurement precision
If multiple trained models processing different data types are used, then defect detection precision is improved, but the system complexity and processing time increase
Solution Approach 1:
The inspection process is segmented into parallel processing streams: the first trained model processes image data while the second trained model processes recording information simultaneously. This segmentation allows both models to operate independently and concurrently, maintaining high detection precision while minimizing processing time through parallel execution
Solution Approach 2:
The results from the first trained model (image-based detection) and the second trained model (recording information-based detection) are merged through a result integration unit. This merging combines the strengths of both models to produce a comprehensive defect detection result, achieving high precision without requiring sequential processing that would slow down inspection speed
Data Source
AI summary
An image inspection system that inspects an image recorded on a recording medium includes one or more hardware processors; and one or more memories storing one or more programs including instructions for storing a first trained model that is generated by machine learning based on learning recorded images that are images for machine learning, recorded on recording media, storing a second trained model that is generated by machine learning based on recording information that is different from the learning recorded images, acquiring a first probability of a defect being in an actual recorded image, acquiring a first estimation result, acquiring a second probability of a defect being in the actual recorded image; acquiring as a second estimation result and detecting a defect in the actual recorded image on the basis of the first estimation result and the second estimation result.