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.
Classifying space targets from debris is critical for radar resource management as well as rapid response during the mid-course phase of space target flight. Due to advances in deep learning techniques, various approaches have been studied to classify space targets by using micro-Doppler signatures. Previous studies have only used micro-Doppler signatures such as spectrogram and cadence velocity diagram (CVD), but in this paper, we propose a method to generate micro-Doppler signatures taking into account the relative incident angle that a radar can obtain during the target tracking process. The AlexNet and ResNet-18 networks, which are representative convolutional neural network architectures, are transfer-learned using two types of datasets constructed using the proposed and conventional signatures to classify six classes of space targets and a debris–cone, rounded cone, cone with empennages, cylinder, curved plate, and square plate. Among the proposed signatures, the spectrogram had lower classification accuracy than the conventional spectrogram, but the classification accuracy increased from 88.97% to 92.11% for CVD. Furthermore, when recalculated not with six classes but simply with only two classes of precessing space targets and tumbling debris, the proposed spectrogram and CVD show the classification accuracy of over 99.82% for both AlexNet and ResNet-18. Specially, for two classes, CVD provided results with higher accuracy than the spectrogram.
For various target tracking applications, it is well known that the Kalman filter is the optimal estimator(in the minimum mean-square sense) to predict and estimate the state(position and/or velocity) of linear dynamical systems driven by Gaussian stochastic noise. In the case of nonlinear systems, Extended Kalman filter(EKF) and/or Unscented Kalman filter(UKF) are widely used, which can be viewed as approximations of the(linear) Kalman filter in the sense of the conditional expectation. However, to implement EKF and UKF, the exact dynamical model information and the statistical information of noise are still required. In this paper, we propose the recurrent neural-network based Kalman filter, where its Kalman gain is obtained via the proposed GRU-LSTM based neural-network framework that does not need the precise model information as well as the noise covariance information. By the proposed neural-network based Kalman filter, the state estimation performance is enhanced in terms of the tracking error, which is verified through various linear and nonlinear tracking problems with incomplete model and statistical covariance information.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.