Hyperspectral Image Dimensionality Reduction via Spatial Anomaly Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Hyperspectral images generate vast amounts of data, leading to latency issues and computational burdens due to existing dimensionality reduction techniques that either compromise important information or fail to adequately reduce data volume for real-time processing and transmission.
Innovation Solution
A method for reducing hyperspectral image dimensionality using an optimized set of basis vectors, which involves decomposing spectral vectors to derive residual vectors, adding basis vectors if magnitudes exceed thresholds, and optimizing the basis vector set to span a subspace with maximum power, allowing for multiple passes to address anomalies and reduce data volume effectively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If existing dimensionality reduction techniques are used to reduce data volume, then data transmission and processing speed improves, but important information is compromised or lost
Solution Approach 1:
The patent transforms the hyperspectral data from spectral domain to spatial domain by identifying and extracting spatial anomalies. This parameter transformation allows the system to reduce data dimensionality while preserving critical spatial information that would be lost in traditional spectral compression methods. The spatial anomaly detection approach changes the representation parameters from wavelength-based to location-based, enabling efficient compression without information loss.
2Reliability
If full hyperspectral data is processed and transmitted, then data integrity is maintained, but processing time and computational burden increase significantly
Solution Approach 1:
The patent extracts only the essential spatial anomaly information from the full hyperspectral dataset, separating critical data from redundant information. By taking out only the spatial anomalies that contain meaningful information, the system maintains data integrity for important features while discarding redundant spectral data, thus reducing processing time without sacrificing reliability.
Solution Approach 2:
The patent segments the hyperspectral processing task into two distinct phases: (1) spatial anomaly detection and identification, and (2) spectral analysis of identified anomalies. This segmentation allows the system to first reduce the data to only relevant spatial locations, then perform detailed spectral analysis only where needed, significantly reducing overall processing time while maintaining complete information integrity for anomaly detection.
3Productivity
If data volume is reduced for real-time processing, then processing speed improves, but the ability to detect anomalies deteriorates
Solution Approach 1:
Instead of the conventional approach of performing spectral analysis first and then spatial analysis, the patent inverts the processing order by performing spatial anomaly detection first, followed by spectral analysis of the detected anomalies. This inversion allows the system to identify all potential anomaly locations in the reduced spatial dataset, ensuring no anomalies are missed, and then apply detailed spectral analysis only to those specific locations, maintaining detection accuracy while enabling real-time processing.
Data Source
AI summary
A method for reducing dimensionality of hyperspectral images may include receiving a hyperspectral image having a plurality of pixels. A basis vector set including a number of members may then be established, wherein each of the members comprises a basis vector. For each of the plurality of pixels, a spectral vector for the pixel may be read and decomposed with the members of the basis vector set to derive a residual vector for the pixel. A basis vector for the pixel may then be added to the members of the basis vector set if the residual vector for the pixel has a magnitude exceeding a predetermined threshold, and the basis vector set may then be optimized to eliminate one of the members of the basis vector set, whereby the optimized basis vector set includes the number of members. A system configured to perform the method may also be provided.