The Brigham Young University Synthetic Aperture Radar (YSAR) is a compact, inexpensive SAR system which can be flown on a small aircraft. The system has exhibited a resolution of approximately 0.8 m by 0.8 m in test flights in calm conditions. YSAR has been used to collect data over archeological sites in Israel. Using a relatively low frequency (2.1 GHz), we hope to be able to identify walls or other archeological features to assist in excavation. A large data set of radar and photographic data have been collected over sites at Tel Safi, Qumran, Tel Micnah, and the Zippori National Forest in Israel. We show sample images from the archeological data. We are currently working on improved autofocus algorithms for this data and are developing a small, low-cost interferometric SAR system (YINSAR) for operation from a small aircraft.

The standard ISAR high-frequency weak-scatterer model is inappropriate to targets with inlets and cavities, and images created under this model assumption often display artifacts associated with these structures. Since inlets and cavities (typically) make a strong contribution to the radar field scattered from aircraft targets, these artifacts often confound the image interpretation process and considerable effort has been spent in recent years to model, isolate, and remove these sources of error. Many of the more complete and accurate scattering models require extensive knowledge about the cavity/inlet shape and size and, moreover, are numerically intensive -- features that make them unsuitable for many imaging applications. We examine an older (and less accurate) model based on a weak-scattering modal expansion of the structure which appears to be well-suited to ISAR imaging. In addition, the analysis shows how cavity/inlet shape-specific information may be estimated from an ordinary ISAR image.

We present a new method for estimating the motion parameters of a target from its inverse synthetic aperture radar (ISAR) signature. This method uses the phase of the target's echo transfer function to calculate a focal quality indicator while avoiding two-dimensional Fourier processing. The focal quality indicator reaches the global minimum of a parametric motion surface when the phase is compensated with the target's actual motion parameters. The presence of an absolute minimum without local minima guarantees that the estimated motion parameters are an accurate representation of the target's motion and allows the use of a simple search procedure. Polynomial fitting is incorporated to the new method to improve the robustness by reducing estimation errors due to the finite order of the parametric motion model.

Inverse-synthetic aperture radar (ISAR) processing algorithms require an explicit or implicit model of the motion of the target, since this information is not measurable at the sensor. A major unsolved problem is whether it is better to develop a deterministic model of the motion or to fit the observations to a mathematical model which may be ambiguous with a wide range of physical target motions. This question is particularly important for the imaging of small targets in ground clutter since the available information on the target motion may be very limited. By contrast, the imaging of large ships in sea clutter yields a wealth of information of the motion of the body. This paper explores the effectiveness of models of the second kind by testing a set of motion estimation algorithms against motor vehicle targets imaged by a high-frequency SAR at long range and low grazing angle. The algorithms attempt to estimate the bulk motion of the vehicle and its rotation about its mean position by fitting many small pieces of information gathered from low quality or short duration scatterers detected in either the raw range- compressed signal history or in the complex image. In forming the motion model, the algorithm avoids undue reliance on particular bright or narrow-Doppler targets in favor of a global fit to many partially resolved or glinting features. The algorithm is demonstrated for 0.3 meter resolution SAR data of a Winnebago van from a Ku-band radar.

The purpose of this article is to describe and compare different numerical methods to reconstruct focused ISAR imagery via interpolation in either range-Doppler or frequency domains. Parameters such as amplitude deviation, image entropy, as well as computational efficiency are used to contrast the different approaches presented. It is shown that conventional linear and cubic interpolation techniques are less accurate than other weighted integration techniques, including the unified Fourier reconstruction algorithm which uses an Airy pattern as the interpolating kernel. The appearance of artifacts in linear and cubic interpolation methods is illustrated and discussed. A point target model of a navy drone is used to compare the effectiveness of each method.

An efficient implementation of the maximum likelihood estimator (MLE) is presented for the estimation of target range, radial velocity and acceleration when the radar waveform consists of a wideband linear frequency modulated (LFM) pulse train. Analytic properties of the associated wideband ambiguity function are derived; in particular the ambiguity function, with acceleration set to zero, is derived in closed form. Convexity and symmetry properties of the ambiguity function over range, velocity and acceleration are presented; these are useful for determining region and speed of convergence for recursive algorithms used to compute the MLE. In addition, the Cramer Rao bound is computed in closed form which shows that the velocity bound is decoupled from the corresponding bounds in range and acceleration. A fast MLE is then proposed which uses the Hough transform (HT) to initialize the MLE algorithm. Monte Carlo simulations show that the MLE attains the Cramer-Rao bound for low to moderate signal-to-noise depending on the a priori estimates of range, velocity and acceleration.

An ultrawideband random noise (white Gaussian noise) radar system operating in the 1 - 2 GHz frequency range has been used to estimate the Doppler spectra of moving objects. A unique technique has been developed to introduce coherence into the system by performing heterodyne correlation of the received signal with the time delayed replica of the transmitted signal. This operation preserves the phase of the reflected signal which is generally lost in traditional homodyne correlation receivers. Knowledge of the phase of the received signal and its time dependence due to the motion of the target permits the system to be configured as a Doppler radar for detecting both linear and rotational motion. This paper describes the basic theory of random noise Doppler radar and presents simulated and experimental results obtained using the University of Nebraska's 1 - 2 GHz random noise radar system.

This paper presents an application of the WVD (Wigner-Ville Distribution) to the problem of focusing an inverse SAR (ISAR) image over an extended period T of the order of two seconds or more. Over short subintervals or segments of duration (Delta) T covering the interval T, subimages may be formed via the normal weighted FFT applied to aligned complex range profiles. These subimages are of low resolution 1/(Delta) T, and though normally in focus, they are of limited use in estimating the parameters governing defocus over the longer interval T. A method is presented here whence subimages may be derived from a Wigner-Ville analysis, which have a resolution of the order of 1/T. This involves no paradox, since coherent data is assumed to be collected over the full duration T. Since the high resolution Wigner subimages contain detailed structures on the more stable scattering centers on the target, a simple correlation analysis can determine defocus parameters with much better accuracy than with conventional subimages, hence a focused image on the entire interval may be determined if desired. However for many purposes the stacked WVD subimages will by themselves provide sufficient information for target identification purposes.

We describe an approach for inverse synthetic aperture radar (ISAR) imaging based on the Gabor wavelet transform. The Gabor basis function introduces three signal parameters which allow the components of a target signature to be mapped into a three-dimensional domain. Time, range, and frequency constitute the dimensions of this domain. Component distribution over the range and frequency dimensions corresponds to a snap shot of the target's scattering centers at a particular observation time. The snap shot resolution can be adjusted in each dimension by properly selecting the Gabor wavelet parameters. Parameter selection, as discussed in this paper, is used to minimize the quadratic phase distortion associated with moving target components. For multiple targets experiencing different velocities, selective motion compensation is incorporated to the Gabor wavelet transform approach, thus yielding focused imagery.

In this paper, we discuss the issue of radar imaging of multiple moving targets. When multiple moving targets are close to each other, the return signals from these targets are overlapped in time. Thus, by applying conventional motion compensation algorithms, when targets have rotational motion or maneuvering, multiple targets may not be resolved and each individual target cannot be clearly imaged. However, in cases where each individual target has its own velocity and moving direction different from others, then the different Doppler histories can be utilized to separate targets from each others. First we describe multiple target resolution. Then, we introduce the time-frequency based phase correction algorithm and discuss its limitation. Based on our previous work on time-frequency image formation, we propose a new method for radar imaging of multiple moving targets. With the new method, targets can be either point or extended targets, and target's motion can be either translational or rotational motion. Some examples are given for comparing the new method with the conventional Fourier method and time-frequency based phase correction methods.

In this paper the chirp z-transform is adapted for the spectral reconstruction of a wideband signal. Here, spectral reconstruction is based on the coherent and sequential processing of non-overlapping short-time segments. This approach eases hardware implementation in the presence of limited memory allocation and can provide savings in computational effort. In addition, it offers the flexibility of focusing the reconstruction of the spectrum on a narrow spectral band of interest.

This paper proposes using a backpropagation (BP) neural network for the classification of ship targets in airborne synthetic aperture radar (SAR) imagery. The ship targets consisted of 2 destroyers, 2 cruisers, 2 aircraft carriers, a frigate and a supply ship. A SAR image simulator was employed to generate a training set, a validation set, and a test set for the BP classifier. The features required for classification were extracted from the SAR imagery using three different methods. The first method used a reduced resolution version of the whole SAR image as input to the BP classifier using simple averaging. The other two methods used the SAR image range profile either before or after a local-statistics noise filtering algorithm for speckle reduction. Performance on an extensive test set demonstrated the performance and computational advantages of applying the neural classification approach to targets in airborne SAR imagery. Improvements due to the use of multi-resolution features were also observed.

This paper describes a fast processing algorithm developed in order to extract sea surfaces' features from radar images. Based on a parametrial estimation, this algorithm allows real time applications by giving a continuous stream of characteristic parameters such as the wavelength and the direction of the current. In addition to its high processing speed, the proposed algorithm, by taking into account the intrinsic polar coordinates of the radar, cancels the effect of distortion introduced into the signal spectrum by a relatively slow radar rotation speed.

Radar images show a characteristic interference, the speckle effect, as a result of the coherent nature of the radar signal, complicating classifications based on the normal distribution. An algorithm is introduced to overcome the disadvantages of a pixel-based classification by use of an object-based approach through the use of digital field boundaries. Besides estimating the characteristic parameters on a field- rather than on a pixel-basis, a method is developed to incorporate prior knowledge into the classification process. An object-based texture approach is compared to a pixel-based approach. Both methods are based on the multinomial distribution. The texture approach incorporates the full information stored in the co-occurrence- matrix in the classification process. The pixel-based approach models the entire histogram over the multinomial distribution. The prior knowledge is formulated in the form of transition matrices. The results of the image classification and the prior knowledge are combined using Dempster's Rule of Shafer's Theory of Evidence because of its ability to combine probability values of various probability distributions. The algorithm was tested with rotation schemes of an agricultural area and, depending on the classification method, showed classification accuracies up to 94 percent.

Statistical and neural network approaches to the classification process of automatic target recognition (ATR) with a synthetic aperture radar (SAR) imaging mode for four ground vehicles are investigated and their performance compared. A set of image features is extracted from a training set of SAR images. A subset of these image features is selected which maximizes the likelihood of correct classification assuming a Gaussian feature distribution. An improved method for statistical classification is demonstrated in which training data is selected based on its statistical variation with azimuth angle. With proper selection of image features it is shown that the misclassification rates of both the statistical and neural network classifiers are approximately the same.

In this paper, we analyze the complexity of an optimal algorithm for realizing the permutation test applied to nonparametric radar detection against the complexity of rank test realization. For a primitive permutation test algorithm, the computational work is very high and its implementation in real-time is difficult, due to the number of operations increases with the number of reference samples (M) to the power of the number of integrated pulses (N) (i.e. M^N). We propose new permutation test and rank test algorithms, and analyze the complexity with respect to N and M for a given false-alarm probability (P_fa); also, the detection probability (P_d) will be evaluated for each case. Optimum values of N and M for a given P_fa will be given for the permutation test and the rank test, resulting in similar values of computational complexity of both of them, i.e. computational complexity proportional to N (DOT) M. We also show the detectability curves of the optimum permutation test versus optimum rank test under Gaussian noise environments for different values of N and M and different target models.

Three-dimensional images free of sidelobe interaction effects of CLEAN have been generated by the PML algorithm. Some implementation speedups are possible and the significance of PML may extend to autofocus. Roll data while traversing a circular path is found suitable to define a curved aperture for preliminary three-dimensional ship imagery.

In phase shifting interferometry, many error-compensating algorithms have been reported. Such algorithms suppress systematic errors caused by nonlinear sensitivities of the phase shifter and nonsinusoidal waveforms of the signal. However, in a Fizeau interferometer where both error sources are equally dominant, the most common group of the algorithms produces errors comparable to those produced by discrete Fourier algorithms which have no capability to compensate for phase-shift errors. It is shown that if an algorithm has an extended immunity to nonlinear phase shift, it can suppress the effects of both error sources simultaneously and yield much smaller errors. When a phase-shifting algorithm is designed to compensate for the systematic phase-shift errors, it becomes more susceptible to random noise. The susceptibility of phase shifting algorithms to random noise is analyzed with respect to their immunity to phase-shift errors. It is shown that for the most common algorithms for nonlinear phase shift, random errors increase as the number of samples becomes large. This class of algorithms has an optimum number of samples for minimizing the random errors, which is not observed in the Fourier algorithms. However, for the new algorithms with an extended immunity to nonlinear phase shift, random errors decrease as the number of samples increases.

Optoelectronic SAR signal processing for real-time parallel adaptive on-board wide area surveillance and ATR applications is described. The signal processing architecture consists of a coarse ATR processor that performs early detection of small targets to select regions of interest in a large field of view, and a fine ATR processor for accurate classification of selected targets. Both coarse and fine ATR processors apply circular correlation-based algorithms implemented with optical joint transform correlators. Together with associative memory and genetic algorithm composite filtering, this discriminates among similar targets and rejects cluttered backgrounds.

We apply the method of ordered multiple interactions (MOMI) to the magnetic field integral equation for a scattering problem involving an infinitely long cylinder suspended over a numerically-generated randomly rough surface. The cylinder and rough surface are infinite conductors. We consider two- dimensional geometry (invariance in the y-direction) and calculate the scattering of a plane wave which is polarized in the plane of scattering. We examine some convergence characteristics of the MOMI method when applied to this scattering problem. Our results indicate that the MOMI method does not converge for scattering only from isolated cylinders, but converges rapidly for scattering only from randomly rough surfaces. To obtain solutions for the problem with both cylinder and rough surface present, we adopt a hybrid method where the cylinder portion is solved exactly and the rough surface portion is solved by iteration.