Lens Cross-Section Identification Using CNN Feature Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods struggle to accurately identify lenses from cross-sectional images due to variations in image quality and drawing styles, requiring extensive rule creation and failing to recognize unfamiliar drawing styles.
Innovation Solution
An information processing apparatus and method using machine learning to detect and identify lenses in cross-sectional images through a convolutional neural network (CNN) and a specific model, constructing a database for lens information and calculating similarity based on feature amounts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a rule-based identification method is used for each drawing style, then the identification accuracy for known styles is improved, but the device complexity and time required to prepare identification rules increase significantly
Solution Approach 1:
The patent replaces the mechanical rule-based identification system with a machine learning-based automated system. The identification model learns patterns from training images and automatically identifies lenses in target images without requiring manual rule creation for each drawing style, thereby reducing device complexity while maintaining or improving identification accuracy.
Solution Approach 2:
The patent changes the approach from fixed rule parameters to adaptive learning parameters. The identification model adjusts its internal parameters through machine learning based on training data, allowing it to handle various drawing styles without manual rule adjustment, thus reducing the complexity of preparing identification rules for different styles.
2Adaptability or versatility
If comprehensive rules for all drawing styles are prepared, then the adaptability to different styles is improved, but the time and effort required to prepare and maintain rules increases
Solution Approach 1:
The patent performs preliminary action by training the identification model in advance using a diverse set of training images representing various drawing styles. This pre-training enables the model to adapt to different styles without requiring additional rule preparation time when encountering new target images, as the adaptability has already been built into the model during the training phase.
Solution Approach 2:
The identification model serves itself by automatically learning and adapting to different drawing styles through machine learning, eliminating the need for human operators to manually prepare and maintain comprehensive rules for all possible styles. The system improves its own adaptability through continuous learning from training data.
3Measurement precision
If manual rule creation for each style is performed, then the identification precision for familiar styles is maintained, but the productivity and efficiency of the identification process decreases
Solution Approach 1:
The patent replaces the manual mechanical process of creating and applying identification rules with an automated machine learning system. The identification model automatically processes target images and identifies lenses without human intervention in rule creation, significantly improving productivity while maintaining identification precision through learned patterns from training data.
Solution Approach 2:
The identification model enables continuous automated processing of target images without interruption for rule creation or adjustment. The machine learning system continuously identifies lenses across different drawing styles using learned patterns, eliminating the discontinuous manual rule preparation process and thereby improving overall identification processing efficiency.
Data Source
AI summary
An object of the present invention is to provide an information processing apparatus, an information processing method, and a program which make it possible to appropriately identify a lens from an image showing a cross section of the lens.In the present invention, a processor detects an existing region of a lens in a target image showing a cross section of a part including the lens in a target device including the lens, and the processor identifies the lens of the target device existing in the existing region based on a feature amount of the existing region by an identification model constructed by machine learning using a plurality of learning images showing a cross section of the lens.