N-Dimensional Image Classification via Geometric Structure Measures
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image classification methods, such as kernel-based segmentation and window-independent classification, often result in scattered classifications and require significant manual work, especially when knowledge of the image content is limited, making it difficult to achieve accurate and reliable automatic classification of images in two or more dimensions.
Innovation Solution
A method that involves identifying a variable geometric structure in an N-dimensional dataset, calculating geometric measures associated with this structure, and using these measures to classify pixels into a main set of classes based on comparative measures, thereby reducing manual effort and improving classification accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If kernel-based segmentation or window-independent classification is used, then automatic classification can be achieved, but the classification result becomes scattered and requires large amounts of manual work
Solution Approach 1:
The patent divides the image into discrete elements and processes each element independently to determine its class identity. This segmentation approach allows automatic classification while maintaining reliability by evaluating each element's geometric structure separately rather than using scattered window-based methods.
Solution Approach 2:
The patent introduces geometric measures as an additional dimension for classification. Instead of relying solely on traditional pixel-based methods, the patent calculates geometric measures (such as area, perimeter, shape factors) of elements to classify them, adding a new dimension of analysis that improves classification reliability while maintaining automation.
2Reliability
If traditional classification methods are used, then processing can be performed, but significant manual work is required to achieve sufficient reliability
Solution Approach 1:
The patent enables the classification system to automatically determine class identities for all elements in the image without requiring manual intervention. Each element's geometric structure is automatically analyzed and compared against reference geometric structures, allowing the system to serve itself and eliminate manual classification work while maintaining high reliability.
Solution Approach 2:
The patent changes the classification parameters from traditional pixel intensity-based methods to geometric measure-based parameters. By calculating and comparing geometric measures (area, perimeter, shape factors) of image elements, the system achieves reliable automatic classification without manual work, as the geometric parameters provide more discriminative power for different land cover types.
3Measurement precision
If geometric measures are calculated for each element, then classification accuracy improves, but computational complexity increases
Solution Approach 1:
The patent extracts only the essential geometric measures needed for classification (such as area, perimeter, and key shape factors) rather than calculating all possible geometric parameters. This extraction approach maintains high classification accuracy by focusing on the most discriminative geometric features while reducing computational complexity by eliminating redundant calculations.
Data Source
AI summary
Method for classifying a two- or higher dimensional image, where each pixel is associated with M property measures, includes identifying firstly a certain predetermined, variable geometric structure, the extension of which in at least two of the N dimensions in the dataset is determined in relation to a single element in the dataset and by at least one variable parameter, and secondly at least one geometric measure associated with the variable geometric structure, which geometric measure is arranged to measure a geometric property of a specific geometric structure in relation to other specific such geometric structures, and in that a main classification is conducted of the dataset, which main classification is based upon a comparative measure between the respective sets of associated geometric measures of two elements, calculated from a respective maximal geometric structure for each element.