Multi-Frequency Sonar Change Detection via SVDD Feature Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current sonar image analysis techniques face challenges in real-time detection and correlation of objects in underwater environments due to the vast quantity of data and high registration requirements for pixel-level processing, making it impractical for humans to identify new objects in sonar images.
Innovation Solution
The use of support vector data description (SVDD) statistics to represent multi-dimensional sonar data, including multi-frequency band data, and characterize objects based on invariant features, allowing for real-time detection and correlation of objects with known objects, and management of a database with confidence metrics for new objects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If pixel-level processing is used to detect changes in sonar data, then measurement precision is improved, but device complexity and loss of time increase due to high registration requirements
Solution Approach 1:
The patent extracts and processes only the most salient features from sonar images rather than performing pixel-level analysis. This involves identifying key characteristics of objects and changes while discarding redundant pixel data, thereby reducing registration complexity while preserving detection precision.
Solution Approach 2:
The patent segments the sonar data processing into distinct stages: initial change detection, feature extraction, and correlation analysis. This segmentation allows the system to avoid the computational burden of complete pixel-level registration while maintaining accurate change detection through strategic feature comparison.
2Measurement precision
If pixel-level processing is used to detect changes in sonar data, then measurement precision is improved, but loss of time increases making real-time detection impractical
Solution Approach 1:
The patent extracts only the essential features needed for change detection rather than analyzing all pixel data. This feature extraction approach maintains measurement precision by focusing on salient characteristics while dramatically reducing processing time to enable real-time operation.
Solution Approach 2:
The patent applies partial action by performing change detection on a subset of critical features rather than complete pixel-level analysis. This selective approach provides sufficient detection precision for real-time applications without the computational overhead of exhaustive pixel comparison.
3Productivity
If automation is increased to process vast quantities of sonar data, then productivity is improved, but difficulty of detecting and measuring increases
Solution Approach 1:
The patent introduces feature extraction as an intermediary step between raw sonar data and object correlation. This intermediary process transforms vast quantities of raw data into manageable feature representations, maintaining high productivity while reducing the difficulty of subsequent correlation and detection operations.
Solution Approach 2:
The patent replaces manual or simple automated pixel-level comparison with sophisticated feature-based processing. This substitution uses advanced algorithms to automatically extract and compare meaningful features, increasing productivity while managing the complexity of detecting and measuring changes through intelligent pattern recognition.
Data Source
AI summary
Methods and systems to detect changes in a region of space relative to baseline measurements, including to process data to detect potential objects, or contacts against a natural background environment, to characterize and geo-register the contacts, to compare and correlate the contacts with a database of known objects, and to report uncorrelated contacts as new objects in the space. Features disclosed herein may be implemented to process image data from a line-by-line image-generation system including, for example, side-looking sonar. Methods and system may be implemented with respect to multi-frequency band sonar data.