Document Image Reading Apparatus Defect Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image reading apparatuses struggle to accurately detect and handle defects in document sheets and written information, particularly when documents are dog-eared or have sticky notes, leading to incomplete or distorted images.
Innovation Solution
The proposed image reading apparatus includes a controller, a sheet defect decider, a written information defect decider, and a decision device. It sequentially reads documents and decides if they have defects in sheet shape and written information using deep learning-based convolutional neural networks (CNNs), allowing for suspension or continuation of the reading process based on user settings.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the image reading apparatus reads all documents sequentially without defect detection, then the reading speed is maintained, but defective documents with dog-ears or sticky notes produce incomplete or distorted images
Solution Approach 1:
The system performs preliminary defect detection by analyzing document images during the sequential reading process. The sheet defect decider identifies physical defects like dog-ears and the written information defect decider detects sticky notes before they affect the final image output, allowing selective handling of defective documents while maintaining overall reading efficiency
Solution Approach 2:
The defect detection system provides feedback to the control device about the quality of each document being read. Based on this feedback, the system can automatically suspend reading of defective documents or alert the user, creating a closed-loop control system that prevents incomplete images from being generated while maintaining high reading speeds for quality documents
2Reliability
If the apparatus suspends reading whenever a defect is detected, then image quality is ensured, but the reading process becomes inefficient due to frequent interruptions
Solution Approach 1:
The system applies different handling strategies to different documents based on their defect status. Quality documents are read continuously without interruption, while only defective documents trigger suspension or user notification. This localized quality control ensures high image quality for acceptable documents while minimizing time loss
Solution Approach 2:
The system performs partial defect detection focusing on critical defects like dog-ears and sticky notes that affect image usability. Rather than analyzing every aspect of each document in detail, the system quickly identifies the presence of major defects, enabling fast decision-making about whether to suspend reading without excessive analysis time
3Measurement precision
If deep learning-based defect detection is implemented, then detection precision is improved, but device complexity increases
Solution Approach 1:
The defect detection system is segmented into specialized modules: a sheet defect decider for detecting physical sheet defects like dog-ears, and a written information defect decider for detecting sticky notes. Each module uses convolutional neural networks optimized for its specific detection task, allowing high precision while managing complexity through functional separation
Solution Approach 2:
The system uses convolutional neural networks that have been pre-trained on large datasets of document images. These pre-trained models serve as copied knowledge from extensive training processes, enabling the relatively simple device structure to achieve high detection accuracy by leveraging pre-processed intelligence from training data
Data Source
AI summary
An image reading apparatus includes an image reading device that reads an image of a document, a controller that causes the image reading device to sequentially read a plurality of documents one by one, a sheet defect decider that decides whether the document that has been read has a defect in sheet shape, on a basis of a document image acquired by the image reading device, a written information defect decider that decides whether information written on the document that has been read has a defect, on a basis of the document image acquired by the image reading device, and a decision device that decides that the document has a defect, when the sheet defect decider decides that the document has a defect in sheet shape, and the written information defect decider also decides that the information written on the document has a defect.