Discriminator Generation for Multi-Size Pattern Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current pattern detection processes are inefficient due to the time-consuming enlargement or reduction of images and calculation of feature quantities for detecting objects of varying sizes, which hinders the acceleration of the detection process.
Innovation Solution
A discriminator generation device that extracts feature quantities from pattern groups of different sizes and generates multiple discriminators for each size, allowing direct detection without image resizing, and includes a second generation unit to create discriminators for intermediate sizes by correcting the determination threshold values, improving discrimination speed and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a single discriminator is applied to images of various sizes by enlarging or reducing images, then detection capability for multiple sizes is achieved, but detection time increases significantly
Solution Approach 1:
The patent segments the detection task by creating multiple discriminators, each specialized for a specific size category (small, medium, large detection objects). Instead of using one discriminator for all sizes, the system divides the detection space into segments and assigns dedicated discriminators to each segment, eliminating the need for image resizing operations.
Solution Approach 2:
The patent changes the parameter of discriminator size to match the size of detection objects. By generating discriminators with different sizes corresponding to small, medium, and large detection objects, the system adapts the discriminator parameters to the detection task, avoiding time-consuming image enlargement or reduction operations.
2Measurement precision
If images are enlarged or reduced to match discriminator size, then detection accuracy is maintained, but processing speed decreases
Solution Approach 1:
The patent performs preliminary action by pre-generating multiple discriminators of different sizes before the actual detection process. This preparation step creates ready-to-use detectors for various object sizes, so during detection, the system can directly apply the appropriate pre-generated discriminator without performing time-consuming image resizing operations.
3Productivity
If multiple discriminators are generated for different sizes, then detection speed increases, but device complexity increases
Solution Approach 1:
The patent adds a size dimension to the discriminator system by creating discriminators with different size parameters. This dimensional expansion allows the system to handle various object sizes efficiently, as each discriminator is optimized for its specific size range, improving detection speed without requiring complex image processing operations.
4Measurement precision
If image resizing is performed for each detection object size, then accurate detection is achieved, but computational overhead increases
Solution Approach 1:
The patent inverts the conventional approach by instead of resizing images to match a fixed discriminator size, it resizes the discriminators to match the image size and object scale. This inversion eliminates the need for image resizing operations, reducing computational overhead while maintaining detection accuracy through size-appropriate discriminators.
Data Source
AI summary
A discriminator generation device 10 includes a feature quantity extraction unit 13 which, using at least two pattern groups having patterns of different sizes for a detection object, extracts a feature quantity of patterns configuring each pattern group, and a discriminator generation unit 14 which generates a discriminator for detecting a detection object of a size corresponding to each pattern group in an image based on the feature quantity of the patterns of each pattern group.