Maritime ATR Data Fusion Using Multiple Neural Classifiers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current Automatic Target Recognition (ATR) systems for radar images, particularly Inverse Synthetic Aperture Radar (ISAR), rely heavily on human intervention and are prone to errors due to the lack of real data characteristics, noise, and inability to handle unknown target types, lacking confidence assessments and accuracy in maritime vessel classification.
Innovation Solution
A system and method for automatic target recognition that combines multiple classifiers, including Gaussian Mixture Model Neural Networks (GMM-NN), Radial Basis Function Neural Networks (RBF-NN), and Vector Quantization (VQ), to extract features from ISAR images, providing confidence estimates and accommodating unknown targets through data fusion and feature vector creation, with a graphical user interface for visual feedback.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multiple classifiers are combined through data fusion, then reliability and accuracy of target recognition is improved, but device complexity increases
Solution Approach 1:
The patent combines multiple classifiers (GMM-NN, RBF-NN, VQ) into a unified data fusion system that integrates their outputs through a classification mapper. This merging approach leverages the strengths of each individual classifier to achieve higher reliability and accuracy in maritime target recognition while maintaining a structured system architecture.
Solution Approach 2:
The classification system is designed to handle multiple target types (surface ships, subs, aircraft) and unknown targets through a universal data fusion framework. The system performs multiple functions including feature extraction, classification, confidence assessment, and adaptation to new target types, making it versatile for various maritime surveillance scenarios.
2Loss of time
If simulated data is used for training, then system development time is reduced, but measurement precision and reliability deteriorate due to lack of real data characteristics
Solution Approach 1:
The system incorporates confidence assessment mechanisms that provide feedback on classification reliability. This feedback loop allows the system to identify when simulated data training may be insufficient and when additional real data training is needed, enabling continuous improvement of measurement precision while managing development time efficiently.
3Ease of operation
If traditional ATR systems are used, then ease of operation is maintained, but adaptability to unknown target types is reduced
Solution Approach 1:
The system employs dynamic adaptation mechanisms that allow it to learn and adjust to new target types in real-time. The confidence assessment and data fusion architecture enable the system to dynamically incorporate unknown targets into its classification framework, maintaining ease of operation while significantly improving adaptability to diverse and evolving maritime threats.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A system and method for performing Automatic Target Recognition by combining the outputs of several classifiers. In one embodiment, feature vectors are extracted from radar images and fed to three classifiers. The classifiers include a Gaussian mixture model neural network, a radial basis function neural network, and a vector quantization classifier. The class designations generated by the classifiers are combined in a weighted voting system, i.e., the mode of the weighted classification decisions is selected as the overall class designation of the target. A confidence metric may be formed from the extent to which the class designations of the several classifiers are the same. This system is also designed to handle unknown target types and subsequent re-integration at a later time, effectively, artificially and automatically increasing the training database size.