Counting Apparatus Using Learning Model for Shape Variation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional counting methods using geometric models are inaccurate when the shapes of count target objects are slightly different, leading to low accuracy in counting.
Innovation Solution
A counting apparatus and learning model producing apparatus that utilize a learning model trained with pairs of training input images and output images to convert count target objects into figures, allowing for accurate counting by acquiring and processing captured images, and incorporating confidence levels to enhance accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If matching processing using a geometric model is performed, then counting can be automated, but accuracy deteriorates when shapes of count target objects are slightly different
Solution Approach 1:
The patent replaces the geometric model matching approach (mechanical/rigid transformation based) with a deep learning-based image processing system. The learning model automatically learns features from training images and generates teaching figures that can adapt to shape variations, eliminating the need for rigid geometric models and their associated transformation parameters.
Solution Approach 2:
The patent changes the fundamental parameters of object representation from geometric models (defined by shape, size, position, orientation) to learned features from training data. The learning model captures essential characteristics through training and generates simplified teaching figures that are more robust to variations in shape parameters.
2Device complexity
If a geometric model is generated automatically, then processing can be simplified, but counting accuracy deteriorates when shapes vary
Solution Approach 1:
The patent performs preliminary learning using training images to generate teaching figures before actual counting. This preliminary action captures the essential features and variations of the count target objects, allowing the system to handle shape variations during counting without complex real-time analysis.
Solution Approach 2:
The patent creates teaching figures that are simplified copies or representations of the count target objects. These teaching figures are generated from training images and capture the essential characteristics, serving as templates for accurate counting even when actual objects have shape variations.
3Measurement precision
If confidence levels are incorporated for each count target figure, then accuracy improves, but processing complexity increases
Solution Approach 1:
The patent incorporates confidence levels as feedback information for each count target figure. The confidence level indicates the reliability of the teaching figure generation and is used to filter or weight counting results, allowing the system to automatically adjust counting accuracy based on the quality of detected figures.
Data Source
AI summary
A counting apparatus includes: a storage unit storing a learning model, the learning model being trained using multiple pairs of training input images each obtained by capturing an image of multiple count target objects with the same shape, and training output images each containing teaching figures that are arranged at respective positions of the multiple count target objects. A captured image acquiring unit acquiring a captured image of multiple count target objects; an output image acquiring unit acquiring an output image in which the count target objects contained in the captured image are converted into count target figures, by applying the captured image to the learning model. A counting unit counting the number of count target objects, using the multiple count target figures contained in the output image; and an output unit outputting the number of count target objects counted by the counting unit.