Image Clustering via Centroid Fusion for Timely Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image clustering methods require multiple iterations and consume significant resources, leading to low timeliness in classifying unsampled or newly-added images.
Innovation Solution
An image clustering method that involves obtaining new images, clustering them to form a first cluster, determining a similar historical cluster, calculating the distance between the two clusters, and generating a target cluster by fusing them based on this distance, thereby eliminating the need for secondary clustering processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple clustering iterations are performed to classify un-sampled or newly-added images, then classification accuracy is improved, but resource consumption increases and clustering timeliness deteriorates
Solution Approach 1:
The patent pre-calculates and stores cluster centroids and similarity metrics during the initial clustering phase. When new images need to be classified, the system retrieves pre-computed cluster information and performs direct similarity comparison instead of re-executing the full clustering algorithm, thereby maintaining classification accuracy while dramatically reducing processing time and resource consumption
Solution Approach 2:
The patent creates simplified representations (centroids) of large image clusters that capture the essential characteristics of each cluster. These centroids serve as compact copies that can be quickly compared against new images using simple distance metrics, avoiding the need to compare new images against every image in the original clusters, thus improving timeliness while preserving classification accuracy
2Measurement precision
If multiple clustering iterations are performed to classify un-sampled or newly-added images, then classification completeness is improved, but device complexity and resource occupation increase
Solution Approach 1:
The patent extracts the essential clustering information (centroids, similarity metrics, and cluster boundaries) from the complex multi-iteration clustering process and stores them as reusable artifacts. This extraction transforms the complex iterative process into a simple retrieval and comparison operation, reducing device complexity while maintaining classification completeness
Solution Approach 2:
The system performs clustering analysis in advance and stores the results in a structured format that enables efficient querying and classification of new images. This preliminary action eliminates the need for repeated complex clustering iterations, simplifying the overall system architecture while ensuring complete classification coverage
3Measurement precision
If traditional secondary clustering is used to classify new images, then clustering accuracy is maintained, but productivity and efficiency deteriorate
Solution Approach 1:
Instead of performing full secondary clustering on new images, the patent uses pre-computed cluster centroids as simplified representations. New images are classified by comparing them against these centroids using efficient distance calculations, which maintains clustering accuracy while dramatically improving productivity by avoiding repeated full clustering operations
Solution Approach 2:
The patent changes the parameter representation from full image data to compressed centroid vectors. This parameter transformation enables efficient similarity computation while preserving the essential clustering information needed for accurate classification, thereby improving clustering efficiency without sacrificing accuracy
Data Source
AI summary
The disclosure provides an image clustering method and an image clustering apparatus. The method includes: obtaining new images, and clustering the new images to obtain a first cluster; determining a historical cluster similar to the first cluster as a second cluster from existing historical clusters; obtaining a distance between the first cluster and the second cluster; and generating a target cluster by fusing the first cluster and the second cluster based on the distance. In the image clustering method, with the image clustering apparatus of the disclosure, secondary clustering processing performed on the existing historical clusters based on newly added images is not required, new and old clusters are directly fused.