Segmented Image Subcategory Refinement Using Pixel Feature Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting changes in spatially similar but temporally different high-definition satellite images face challenges in image alignment and refined categorization, particularly due to distortions and the computational expense of identifying sub-categories.
Innovation Solution
A system and method that includes image alignment using similarity searches to identify anchor blocks, followed by refined categorization through pixel-level feature set extraction and similarity matching with known sub-category databases to assign sub-categories to segmented image blocks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image alignment and refined categorization are performed using traditional methods, then categorization accuracy can be improved, but computational complexity and processing time increase significantly
Solution Approach 1:
The patent divides the satellite image into multiple blocks and processes each block independently through the CNN decoder. This segmentation allows parallel processing of different image regions, reducing overall computational complexity while maintaining categorization accuracy for each segment.
Solution Approach 2:
The patent performs image alignment and anchor block identification before the main categorization process. By pre-aligning images and identifying anchor blocks that represent stable features, the system reduces the computational burden during subsequent categorization operations, as these pre-processed elements can be reused across multiple comparisons.
2Measurement precision
If traditional image processing methods are used to identify sub-categories, then categorization detail can be improved, but processing speed decreases
Solution Approach 1:
The patent replaces traditional mechanical image processing methods with a neural network-based CNN decoder. This substitution enables the system to perform complex sub-category identification through learned feature representations rather than explicit algorithmic processing, significantly improving processing speed while maintaining or enhancing accuracy.
Solution Approach 2:
The patent changes the parameter space by working with feature sets extracted from image blocks rather than raw pixel data. By operating in this transformed parameter space, the CNN decoder can efficiently identify sub-categories through similarity comparisons, reducing processing time while maintaining detailed categorization capability.
3Measurement precision
If detailed pixel-level analysis is performed for categorization, then categorization precision can be improved, but computational resources required increase
Solution Approach 1:
The patent extracts relevant feature sets from image blocks using the CNN encoder before passing them to the decoder for categorization. This extraction process separates the essential categorical features from redundant pixel-level information, allowing detailed categorization to be performed on compressed feature representations that require fewer computational resources.
Solution Approach 2:
The patent transforms the problem from operating in pixel space to operating in feature space through the CNN encoding process. This dimensional transformation allows the system to achieve detailed categorization precision by working with lower-dimensional feature vectors that capture essential information more efficiently, reducing the computational resources needed for analysis.
Data Source
AI summary
A system to refine segmented image categories includes a pixel feature set extractor, a known sub-category database, a pixel feature set searcher and a sub-category assignor. The extractor extracts pixel level feature sets corresponding to segmented image blocks, the segmented image blocks having data and category metadata. The known sub-category database stores known sub-category feature sets extracted from segmented images with known sub-categories. The pixel feature set searcher matches query pixel feature sets to candidate known sub-category feature sets using a similarity search and the sub-category assigner adds sub-category metadata to the segmented image block metadata.