Search

Traditional synthetic aperture radar (SAR) systems tend to discard phase information of formed complex radar imagery prior to automatic target recognition (ATR). This practice was historically driven by available hardware storage, processing capabilities, and data link capacity. Here, we characterize the value of phase in SAR imagery when estimating target features utilized in ATR algorithms for target discrimination.

In a radar environment, cognitive receivers typically process raw IQ samples into sequences of pulse descriptor words (PDWs). When there are multiple emitters present in the signal, their respective PDWs are interleaved in time, necessitating specialized signal processing to deinterleave the pulse trains. Many deinterleaving solutions in the literature today utilize clustering algorithms to summarize the potentially massive amount of PDW data. Often, these clustering algorithms do not account for the fact that the number of clusters can change as a function of time. Additionally, PDW data arrives in a continuous stream, necessitating an efficient implementation. In this paper, we use Bayesian nonparametrics to address a dynamically changing radar scenario and implement an efficient online algorithm. We show that this algorithm can provide an effective solution to radar pulse clustering.

We extend the concept of monostatic aspect diversity for improved automatic target recognition (ATR) to a bistatic synthetic aperture radar (SAR) platform. We derive the probability of error with respect to the number of aspects used for a simple two-target template-matching classification system. The validity of the error prediction is confirmed using simulated bistatic SAR images. Our results demonstrate the ATR benefits of supplementing monostatic SAR images with one or more bistatic SAR images.

In this paper, a novel algorithm, SV-Means, is formulated for open set radar waveform classification. The approach extends the quantile one-class support vector machine (q-OCSVM) density estimation algorithm into a classification formulation with inspiration from k-means and stochastic gradient descent principles. Phase-modulated radar waveform data is used to compare the primal and dual q-OCSVM (configured for classification) to SV-Means to verify the algorithm's classification and rejection performance. SV-Means is also shown to be an effective open set classification algorithm and is compared to these other open set classifiers: 1-vs-set machine, W-SVM, PI-SVM, and probabilistic open space (POS)-SVM. The SV-Means algorithm proves to achieve similar, and some cases better, performance against comparable algorithms in a fraction of the training time (as the data records increase and, in the case of the q-OCSVM, level sets increase). The SV-Means open set algorithm is an attractive adaptive classification framework using any feature extraction approach including deep-learning architectures.

An important concern in radar is understanding the consequence of waveform design constraints on the signal-to-interference-plus-noise ratio (SINR) performance. In this study, the authors develop performance models for the attendant drop in SINR as an integrated sidelobe (ISL) constraint is placed on the waveform. The performance models are developed using an eigenbasis subspace approach. The authors validate the performance models against measured data and conclude that the proposed approach provides reasonable performance modelling of the SINR as a function of the ISL constraint.

In this paper, the problem of moving target detection (MTD) in distributed multiple-input multiple-output (MIMO) radar in non-homogeneous environments is addressed. The MTD is performed in the Doppler domain and the received signal disturbance is modeled using a linear combination of multiple discrete clutter scatterers whose Doppler frequencies and powers vary from one range bin to another and from one transmit- receive pair to another. We develop a computationally efficient algorithm for MTD using fast iterative interpolated beamforming (FIIB) techniques. Unlike existing MTD techniques which apply orthogonal projections to eliminate powerful clutter discretes from the cell under test, the proposed method mitigates the clutter in the frequency domain via spectral leakage elimination. The computational complexity of the proposed method is shown to be linearly proportional to the computational complexity of the fast Fourier transform, the number of iterations of the FIIB algorithm, and the number of discrete clutter components. It is also shown that the detection performance of the proposed method is similar to that of orthogonal projection based techniques, at a reduced computational cost. Simulation examples are provided to demonstrate the effectiveness of the proposed method.

In this paper, we consider the problem of moving target detection (MTD) in distributed multiple-input multiple- output (MIMO) radar in non-homogeneous environments with robustness against clutter mismatch. The non-homogeneous clutter is modeled in the Doppler domian as a linear combination of multiple discrete scatterers whose Doppler frequencies and powers vary from range bin to another and from transmit- receive pair to another. We develop a new method for MTD with robustness against mismatch between the actual and presumed discrete clutter Doppler frequencies. The proposed method first obtains an accurate estimate of the non-homogeneous clutter Doppler frequencies for every transmit-receive pair in the system using computationally-efficient fast Fourier transform (FFT)- based techniques. Then, the method mitigates the heterogeneous clutter by projecting the received data onto the orthogonal subspace of the discrete clutter, followed by the application of generalized-likelihood ratio test (GLRT) detection. Simulation examples are provided to demonstrate the robustness of the proposed GLRT over existing techniques.

1.Prerequisite material: basics of radar a)Geometric relationships b)Basic radar physics c)Basic range-Doppler processing or imaging 2.Part I - Prof. Rigling a)Introduction to Bistatic Radar b)Bistatic Synthetic Aperture Radar c)Multistatic Synthetic Aperture Radar 3.Part II - Prof. Martorella a)Bistatic Inverse Synthetic Aperture Radar b)Multistatic Inverse Synthetic Aperture Radar

To fully appreciate the benefits of arbitrary waveform design capability for transmit adaptive systems, the trade space between constraints (employed to increase the measure of practicality for radar) and the usual performance driver (signal-to-interference-plus-noise ratio) needs to be better defined and understood.We address this issue by developing performance models for radar waveforms with cumulative-modulus and energy constraints. Radar waveforms typically require a constant-modulus (constant-amplitude) transmit signal to efficiently exploit the available transmit power. However, recent hardware advances and the capability for arbitrary (phase and amplitude) designed waveforms have forced a reexamination of this assumption in order to quantify the impact of the nonconstant modulus property.We develop performance models for the signal-to-interference-plus-noise ratio as a function of the cumulative modulus for a random colored interference environment and validate the models against measured data.

Synthetic aperture radar (SAR) is a form of remote sensing where coherent radar echoes transmitted from a moving platform are processed to form an image of a scene, usually on the ground. There are several algorithms that have been developed with varying levels of complexity and accuracy. In applications with large scene size requirements, the choice of image formation algorithm is important. Exact imaging algorithms like the back-projection algorithm (BPA) can form large images without errors, but they are computationally expensive. Another well-known algorithm is the polar format algorithm (PFA), which is significantly faster than BPA, but it uses approximations that cause image errors in large scenes. In this paper, we evaluate the scene size limitations of the PFA in terms of image defocus. This is caused by residual quadratic phase errors that arise due to approximations in the algorithm. We derive this residual quadratic phase error using a Taylor series expansion in the slow time dimension. Then, we derive simplified expressions for image defocus for two flight paths: circular and linear. We also use the Taylor series expansion to derive accurate corrections for image distortion caused by PFA. These distortion corrections are used in conjunction with the residual quadratic phase errors to derive accurate scene size limitations that are notably different from the circular regions of focus determined in earlier works.

Automatic target recognition (ATR) of synthetic aperture radar (SAR) target chips is a difficult problem, complicated by the number of nuisance parameters typically present in SAR imagery. This also complicates performance analysis because fully sampling the space of nuisance parameters in an evaluation dataset is intractable. ATR algorithms that first quantize pixel intensity values have been shown to be effective for SAR ATR due to hypothetically reducing the sensitivity to these nuisance parameters. Here we study the performance of two such algorithms, multinomial pattern matching and quantized grayscale matching, and compare them with the traditional mean squared error (MSE) template matching based classification approach. Our approach is to approximate the decision statistic of each algorithm as a Gaussian random variable (RV) parameterized by the noise power, or alternatively signal-to-noise ratio (SNR). This allows the analytic prediction of algorithm performance when the noise process of test images differs from that of the dataset used to train each algorithm, without having to rely on costly empirical simulation. We verify our results in simulations utilizing the AFRL "Civilian Vehicle" dataset.

In this chapter we discuss design techniques and constraints that facilitate a waveform that is useful for radar. While signal-to-noise ratio (SNR) and signal-to-interference-plus-noise ratio (SINR) have historically been the primary drivers of waveform design and are certainly discussed here, the growing need for spectral coexistence is eliciting new metrics and design approaches that address a broader set of considerations. This chapter also examines the use of performance prediction models as a means to establish relationships between waveform constraints and SINR performance, which are critical for detection in spectrally crowded environments. An empirical approach is discussed along with models that provide an intuitive way to understand the complex relationships between constraints and SINR performance in dynamic situations.

There is a presumed adverse effect on radar operation brought about from embedding communication signals in radar waveforms. Contrary to that presumption, we show in this paper that modulation of the radar pulse can indeed benefit both radar and communications in the dual system paradigm. This is shown by analyzing the impact of phase-shift keying (PSK) symbol embedding on the ambiguity function (AF) of a multiple-input multiple-output (MIMO) radar. Our analysis shows that the embedded PSK symbols yield reduction in the AF sidelobe peaks for a single pulse as well as for a series of radar pulses. To prove these results, we analyze the AF with and without the PSK symbol embedding. It is shown that the embedding of PSK symbols enable increasing the number of orthogonal transmit waveforms that can be used without increasing the sidelobe levels (SLL) of the corresponding AF.

In this paper, we consider the problem of moving target detection in distributed multiple-input multiple-output (MIMO) radar under the assumption of non-homogeneous environment. In particular, we assume the spatial heterogeneous clutter, i.e., clutter reflections from different range bins to be independent and the clutter power changes from one cell to another. Even under the same test cell, the clutter components associated with different transmit-receive pairs are dissimilar due to the azimuth-selective back-scattering of the non-homogeneous clutter. A new method for estimating the non-homogeneous clutter for every transmit-receive pair in a distributed MIMO radar is developed. For each transmit-receive path, the nonhomogeneous clutter is modeled as a linear combination of appropriately selected basis vectors. The clutter component is adaptively estimated based on maximizing the clutter power and maximally rejecting the power associated with target reflections. A generalized-likelihood ratio test (GLRT) for moving target detection is also developed. Simulations are provided to demonstrate the performance improvement of the proposed GLRT over existing techniques.

The objective of this research is to model anisoplanatic atmospheric effects for synthetically generated Synthetic Aperture Ladar (SAL) data, and to evaluate the impact of these effects on target classification in SAL imagery. Although SAL wavelengths present key advantages in target exploitation, they are heavily susceptible to atmo- spheric turbulence; hence the inclusion of atmospheric effects in simulation is crucial. With utilization of an atmospheric modeling toolbox, we introduce the ability to model anisoplanatic atmospheric effects into a ray tracing simulation tool. We show image classification results as a function of atmospheric severity.

Bistatic Synthetic aperture radar (SAR) Automatic target recognition (ATR) is directly related to how similar targets appear at a given bistatic angle. We investigate the correlation coefficient between bistatic SAR images of opposing targets at various bistatic angles to analyze the effect of bistatic angle on ATR. Results using civilian vehicle targets and an EM simulator demonstrate the value of bistatic perspectives in discriminating between targets. We find that targets often appear less similar at bistatic perspectives than they do at the corresponding monostatic perspective.

In this paper, the L 1 LMS algorithm is considered as an alternative to conventional cross-correlation processing for generating an estimate of a radar range profile. The update equation for the Li LMS filter tap weight vector is derived to enhance the reader's understanding. Results achieved in simulation using MATLAB® are shown, which illustrate robust performance achieved by an LMS-type algorithm employing a regularization term in the cost function to be minimized when processing simulated radar data for which the transmit noise waveform's spectrum contains multiple notches.