Feature Quantity Extraction via Linear Region Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for extracting feature quantities from images, such as those of human skin or other targets, face excessive processing loads, leading to inadequate feature extraction and identification, particularly when dealing with two-dimensional images.
Innovation Solution
A feature quantity extracting device that acquires images, specifies linear regions extending in multiple directions within the image, and extracts feature quantities based on pixel values along these specified regions, reducing the processing load while ensuring accurate extraction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If feature quantities are extracted from the overall captured image and the cropped image, then the identification accuracy is improved, but the processing load becomes excessive
Solution Approach 1:
The patent segments the two-dimensional image data into multiple one-dimensional linear regions extending in different directions. Instead of processing the entire 2D image, the system divides it into several 1D lines, extracting feature quantities from each line separately. This segmentation reduces the overall processing load while maintaining comprehensive coverage of the diagnostic target through multi-directional sampling.
Solution Approach 2:
The patent transforms two-dimensional image data into one-dimensional data by extracting linear regions in multiple directions. This dimensionality reduction from 2D to 1D simplifies the data structure and reduces processing complexity while preserving essential diagnostic information through multi-angular observation of the target.
2Productivity
If feature quantities are extracted from specified regions only, then the processing load is reduced, but the ability to appropriately extract feature quantity for the diagnostic target deteriorates
Solution Approach 1:
The patent applies local quality by extracting linear regions from multiple different directions rather than using a single uniform approach. Each direction provides a different local perspective of the diagnostic target, and the combination of these multi-directional local views ensures comprehensive feature extraction while maintaining reduced processing load compared to full 2D analysis.
3Loss of information
If two-dimensional image data is used for feature extraction, then comprehensive information is obtained, but the extraction load becomes excessive
Solution Approach 1:
The patent reduces information loss by compensating for the 2D-to-1D transformation through multi-directional sampling. By extracting linear regions in multiple directions, the system captures comprehensive information about the diagnostic target from different angles, ensuring that no critical features are lost despite the dimensionality reduction that lowers processing load.
Solution Approach 2:
The patent segments the comprehensive 2D information into multiple 1D linear components distributed in different directions. This segmentation strategy preserves the completeness of information by ensuring that various parts of the diagnostic target are sampled from multiple perspectives, while the 1D format keeps extraction load manageable.
Data Source
AI summary
An identification device includes an acquirer for acquiring a captured image obtained by imaging a predetermined target, a specifier for specifying a linear region extending in a plurality of mutually different directions, from a region of the captured image acquired by the acquirer, the region including the predetermined target, and an extractor for extracting a feature quantity based on the pixel value, from the linear region specified by the specifier.