Blind Hyperspectral Sensing for Clutter Rejection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing surveillance systems face challenges in separating signals of interest from strong background clutter and interference, particularly in environments where signals are weak and buried, and struggle with overlapping signals in space and wavelength, leading to inefficiencies and limitations in signal detection and classification.
Innovation Solution
A blind sensing system that uses processors and non-transitory computer-readable media to capture hyperspectral data, form signal mixtures in orthogonal dimensions, and perform independent component analysis for demixing, allowing for the detection and classification of spectral signatures without prior knowledge of the signals, using techniques like nearest neighbor or neural network classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If prior knowledge filtering is used to extract signals from clutter, then signal detection capability is improved, but the system cannot handle new signals that have never been seen before and cannot easily handle signals overlapping in space and wavelength
Solution Approach 1:
The system performs blind source separation without requiring prior knowledge of signal characteristics. The algorithm automatically adapts to separate signals from clutter by exploiting statistical independence properties, enabling the system to detect both known and novel signals without manual configuration or training data
Solution Approach 2:
The system transforms the signal separation problem from the time domain to the frequency domain, where statistical independence properties become more pronounced. By operating in the frequency domain, the blind source separation algorithm can effectively separate overlapping signals that are mixed in the time domain
2Measurement precision
If angle of arrival separation using phased array antennas is used, then signal and clutter can be separated in angle, but the system is expensive and heavy compared to single element sensors
Solution Approach 1:
The system replaces the mechanical phased array antenna system with a single-element sensor combined with blind source separation signal processing. Instead of using multiple physical elements to achieve spatial separation, the system uses a single sensor element and separates signals computationally in the frequency domain based on statistical independence
Solution Approach 2:
The blind source separation algorithm acts as an intermediary that processes the raw sensor output to separate mixed signals. The algorithm serves as a computational mediator that transforms the single sensor's mixed signal output into separated source signals without requiring direct physical separation at the sensor level
3Measurement precision
If existing hyperspectral ICA demixing methods treat hyperspectral data as a set of images with one image per spectral band, then signal separation can be performed, but the process is slow and inefficient especially when some signal components are much stronger than others
Solution Approach 1:
The system changes the approach from treating each spectral band as a separate 2D image to treating the hyperspectral data as a 4D tensor (two spatial dimensions, one spectral dimension, and one temporal dimension). This dimensional reorganization allows for more efficient batch processing and exploitation of temporal correlations across multiple time points
Solution Approach 2:
The system segments the hyperspectral data processing into distinct stages: forming signal mixtures in some dimensions, demixing in orthogonal dimensions using blind source separation, and then detecting spectral signatures. This segmentation allows each stage to be optimized independently, improving overall processing efficiency
Data Source
AI summary
Described is a blind sensing system for hyperspectral surveillance. During operation, hyperspectral data is captured using a hyperspectral camera as mounted on a mobile platform. The system then forms a signal mixture of a plurality of multi-dimensional signals. The multi-dimensional signals being the captured hyperspectral data of a wide area having a background and an object. The plurality of multi-dimensional signals are then demixed using blind source separation, resulting in separated spectra. Finally, the system detects and recognizes a spectral signature of the object in the separated spectra.