Image Classification via Segmented Weighted Integration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High-resolution image classification using machine learning or deep learning algorithms faces challenges due to performance constraints, leading to reduced accuracy in classification results, as existing methods often rely on segmenting images into small parts, which can misrepresent the entire image's features.
Innovation Solution
An image classification apparatus and method that includes an image segmentation module, a primary classification module, and a secondary classification module, where the primary module performs machine learning on segmented images and the secondary module calculates an optimum weight value combination to enhance classification accuracy by adjusting learning parameters and loss thresholds.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-resolution images are used directly for classification, then classification accuracy should be improved, but performance constraints prevent effective use
Solution Approach 1:
The patent divides high-resolution images into multiple segment images (e.g., 9 segments) to enable processing within performance constraints. Each segment is classified independently by a primary classifier, and the results are integrated by a secondary classifier to achieve accurate classification of the original high-resolution image without overwhelming computational resources
2Productivity
If images are segmented into small parts for classification, then processing performance is improved, but classification accuracy deteriorates due to loss of global image context
Solution Approach 1:
The patent merges the classification results of multiple segment images through a secondary classifier to produce the final classification of the original image. The secondary classifier integrates the primary classification estimation values from all segments, reconstructing the global image context and ensuring that the final classification reflects the entire image rather than just individual segments
3Speed
If segment images are used for classification, then processing speed is improved, but classification results may be determined by feature quantity rather than feature presence
Solution Approach 1:
The patent introduces a secondary classifier as an intermediary between the segment classification results and the final classification decision. The secondary classifier receives primary classification estimation values from all segments and performs weighted integration, ensuring that the final classification is based on the presence and significance of features across the entire image rather than merely the quantity of features in individual segments
Data Source
AI summary
An image classification apparatus includes an image segmentation module configured to segment a learning image into a plurality of segment images, a primary classification module configured to perform machine learning on a primary classifier using the plurality of segment images, and a secondary classification module configured to calculate a weight value combination for creating a secondary classification estimation value for the learning image from a plurality of primary classification estimation values generated by passing the plurality of segment images to the trained primary classifier, or a machine learning-based learning parameter.