Abstract
This study focuses on the reentry target classification and fuses target features based on the generalized evidence theory. The features are extensively investigated, and the ballistic factor and length of the high-resolution range profile are selected. The evidence theory is advantageous for solving feature fusion, representing uncertainty, and is widely used in defense applications. However, determining the generalized basic probability assignment (GBPA) and dealing with uncertainty is a matter that requires further improvement. In this paper, we propose a new method to determine GBPA using uncertainty with time-series radar data. First, the samples of each known class are encoded as a generalized fuzzy number (GFN), and the power set comprising the frame of discernment (FOD) is calculated from the GFN and each intersection area. Subsequently, the test samples with uncertainty are encoded as triangular fuzzy numbers, reflecting the mean and standard deviation of a Kalman filter. Finally, the firing strength between the model and the input is calculated as the degree of support for the class hypothesis, which is used to determine the GBPA. The proposed algorithm is compared with the existing methods and exhibits high classification accuracy and a short classification time without leakage. In experiments with various input uncertainties, the results demonstrate that our method can effectively reflect the input uncertainty and determine the GBPA.
Introduction
In Ballistic Missile Defense (BMD), the reentry phase represents a critical "final countdown" where interceptors must distinguish between lethal warheads and various forms of "clutter," such as spent booster stages or debris, within seconds. This classification task is complicated by inherent radar noise and the extreme speeds of reentry vehicles, which limit the time available for complex feature extraction. Traditional methods often treat sensor data as deterministic points, failing to account for the statistical uncertainty provided by tracking filters. Moreover, these systems frequently struggle with "Open World" scenarios—encountering target types not present in the training data.
Methodology
The proposed approach leverages Generalized Evidence Theory (GET) combined with Fuzzy Logic to determine the Generalized Basic Probability Assignment (GBPA) from time-series radar data.
Feature Selection
The system utilizes two primary features:
- Ballistic Factor (BF): A measure of how much an object decelerates due to atmospheric drag.
- Target Length: Derived from High-Resolution Range Profiles (HRRP).
Fuzzy Encoding and Fusion
A core innovation of this research is the encoding of data as fuzzy numbers to represent uncertainty. Known classes are modeled as Generalized Fuzzy Numbers (GFN). Real-time data from a Kalman filter is then encoded as Triangular Fuzzy Numbers (TFN), where the triangle's shape reflects both the mean and the standard deviation of the radar measurements.
By calculating the "firing strength" between the model and the input, the system determines the degree of support for various class hypotheses. GET is then used to fuse these features, offering superior performance in handling conflicting evidence compared to standard Dempster-Shafer theory.
Experimental Results
The study utilized a simulated dataset of 6,462 instances across 45 trajectory scenarios, including Warheads (W), Boosters (B), and Debris (D). To evaluate "Open World" robustness, "Heavy Decoys" were introduced as unknown targets.
The hybrid fuzzy-evidence model significantly outperformed standard machine learning classifiers:
- Accuracy: The proposed method achieved 94.2% accuracy, surpassing Support Vector Machines (86.5%) and Naive Bayes (88.2%).
- Decision Speed: The system correctly classified warheads in 8.83 seconds, nearly a second faster than SVM (9.68s).
- Reliability: Crucially, the system achieved "no leakage," meaning no warhead was ever misclassified as harmless debris.
Discussion and Conclusion
The integration of uncertainty as a "fuzzy boundary" rather than noise to be removed prevents overconfident, incorrect classifications. The use of a conflict coefficient ($K$) allows the system to identify unknown targets (like heavy decoys) when evidence from different features is contradictory.
While the system shows high promise, future work will focus on addressing sophisticated decoys designed to mimic warhead ballistic coefficients and expanding the dataset to include more diverse atmospheric conditions. This approach provides a flexible, high-performance upgrade for existing defense systems by effectively mapping sensor uncertainty into the classification process.
Summary updated