Search

The radar cross section (RCS) is an important parameter that reflects the scattering characteristics of radar targets. Based on the monostatic radar RCS time series' statistical features by sliding window segmentation, a novel sliding window‐statistical‐gated recurrent unit (SW‐S‐GRU) method for radar target recognition (RTR) is proposed using a GRU. The sliding window method divides the RCS time series into many segments for obtaining the statistical features as the input of the GRU. The deeper features are extracted from the segmented time series and then classified by the GRU. Under the model of precession motion, simulation results show that the proposed method achieves a great recognition accuracy. Even under the condition of a low SNR, the proposed method can obtain better recognition accuracy. For the precession dataset, the effect of sliding window parameters on RTR is analysed. The simulation results show that the recognition result of the model is better using the longer series overlap length between adjacent times. And when the number of hidden layers is 8 or above 8, the recognition accuracy does not change. Using the publicly available measured drone RCS data, the proposed method can also identify the drone well.

Frequency agile radar (FAR) systems have been demonstrated to have outstanding performance in electronic counter‐countermeasures (ECCM). With increased bandwidth in which FAR can switch carrier frequency, the phase errorswill appear at the corresponding frequency point, resulting in the decrease of coherence of the system. To accomplish phase error correction and range and velocity information extraction of the target simultaneously, a joint phase error estimation and sparse scene restoration algorithm based on sparse recovery (JPSA‐SR) is proposed here. To further reduce the computational complexity of the algorithm, a joint phase error estimation and sparse scene restoration algorithm based on alternating direction method of multipliers (JPSA‐ADMM) is also proposed. Numerical examples show the effectiveness of the proposed algorithms.

The target classification of unknown classes using radar sensor data is discussed. The neural network‐based classifier shows high classification accuracy for the learned class targets. However, there is a risk of false decision for the untrained class target owing to an overconfidence problem. The output confidence of the classifier is calibrated using the deep convolutional neural network ensemble structure to propose a method to set the proper threshold for output confidence to decide unknown class targets. When using the proposed method, the accuracy of the learned target is maintained similar to that of the existing single neural network‐based classifier, whereas the unknown class target is better identified. Further analysis verifies the effectiveness of the proposed method using commercial automotive radar. The proposed method can classify learned targets with an accuracy of 95% and distinguish unknown class targets with an accuracy of at least 85%. Based on the interaction with other sensors, individual sensors need to make reserved decisions about uncertain information. It is expected that the proposed ensemble network will be efficient in designing the classifier to perform target classification including unknown class decision.

Narrowband and broadband indoor radar images significantly deteriorate in the presence of target‐dependent and target‐independent static and dynamic clutter arising from walls. A stacked and sparse denoising autoencoder (StackedSDAE) is proposed for mitigating the wall clutter in indoor radar images. The algorithm relies on the availability of clean images and the corresponding noisy images during training and requires no additional information regarding the wall characteristics. The algorithm is evaluated on simulated Doppler‐time spectrograms and high‐range resolution profiles generated for diverse radar frequencies and wall characteristics in around‐the‐corner radar (ACR) scenarios. Additional experiments are performed on range‐enhanced frontal images generated from measurements gathered from a wideband radio frequency imaging sensor. The results from the experiments show that the StackedSDAE successfully reconstructs images that closely resemble those that would be obtained in free space conditions. Furthermore, the incorporation of sparsity and depth in the hidden layer representations within the autoencoder makes the algorithm more robust to low signal‐to‐noise ratio (SNR) and label mismatch between clean and corrupt data during training than the conventional single‐layer DAE. For example, the denoised ACR signatures show a structural similarity above 0.75 to clean free space images at SNR of −10 dB and label mismatch error of 50%.

Bistatic airborne radar has excellent performance due to the separate transmitter and receiver, but its clutter distribution is usually non‐stationary, which causes the performance of the traditional Space time adaptive processing (STAP) method to fail. Herein, firstly a bistatic airborne radar clutter signal model established, followed by an analysis of its clutter distribution characteristics in detail, and finally a bistatic airborne radar space–time compensation STAP method based on adaptive segmentation is proposed. The method first uses the distribution characteristics of the peak spectral centre in the first pulse data of the airborne radar to adaptively segment the clutter data, including extremely non‐stationary, severely non‐stationary, and non‐stationary clutter regions. Then the clutter data are compensated based on the estimation value of the space–time peak spectrum centre. Finally, each range section is processed by using STAP, and the detected target Doppler frequency is adaptively moved. Simulation results verify the effectiveness of the proposed method.

A deep‐learning‐based approach to estimating the number of coherent sources in radar is presented. A proper estimate of the number of sources in a signal enables improved angle‐of‐arrival (AoA) estimation common in applications such as radar, sonar, and communication systems. Many AoA estimators utilised in these areas require accurate estimates of the number of sources for enhanced performance. Herein, a robust method that performs well under the existence of coherent sources is developed. The proposed method is based on deep learning and it is shown to perform better than state‐of‐the‐art versions of the Akaike Information Criteria (AIC), Minimum Description Length (MDL), and Exponentially Embedded Families (EEF) estimators, which employ spatial smoothing of the covariance matrix to handle coherent signals.

Compared with conventional phased array radar, the collocated multiple-input multiple-out (C-MIMO) radar is envisioned to offer greater freedom in waveform design and can simultaneously manage different beams to track multiple targets in the simultaneous multi-beam (SM) working mode. In this study, aiming at the multiple targets tracking (MTT) problem, a joint beam selection integrated with power and bandwidth allocation (JSPBA) scheme for the C-MIMO radar is proposed in the SM mode. By incorporating the modified particle filter, the predicted conditional Cramer–Rao lower bound (PC-CRLB) is calculated, which gives a measure of the best achievable performance for targets tracking. Then, the optimisation model is established with the aim of improving the worst case of PC-CRLB to achieve better performance of the worst-case tracking. Next, a three-step solution method is proposed to solve the JSPBA problem by converting the non-convex problem into a series of convex problems. At last, after the solutions are fed back to the control centre to guide the beam generation in the next tracking epoch, a cognitive tracking system is established. Simulation results confirm the superiority of the proposed method in improving the MTT performance in C-MIMO radar system.

Human identification plays a vital role in daily lives. A majority of biometric technologies require the active cooperation of humans, while gait recognition does not. Compared with other identification technologies, radar-based technology can monitor the human body around the clock without being affected by light/weather, and is not easy to be forged while protecting privacy. Previous researches have revealed that gait signatures acquired using radar can be used for human identification, but there is almost no literature on the long-term stability of gait signatures. Due to the long-term interval observation, the human micro-Doppler will change according to the subject (such as slight differences in walking posture). In this study, a novel network is proposed to realise stable identification of humans by extracting long-term stable features. The micro-Doppler data is processed by a short-time Fourier transform and finally classified by the proposed neural network. Data acquisition was carried out within more than a month. The experimental results demonstrate that the recognition accuracy of the validation set can reach about 99%, and the recognition accuracy of the test set can reach 90% (improved 3% compared with the network without a recurrent neural network), showing the potential of the proposed method in long-term stable identification.

For high-resolution scene mapping and object recognition, optical technologies such as cameras and LiDAR are the sensors of choice. However, for future vehicle autonomy and driver assistance in adverse weather conditions, improvements in automotive radar technology and the development of algorithms and machine learning for robust mapping and recognition are essential. In this study, the authors describe a methodology based on deep neural networks to recognise objects in 300 GHz radar images using the returned power data only, investigating robustness to changes in range, orientation and different receivers in a laboratory environment. As the training data is limited, they have also investigated the effects of transfer learning. As a necessary first step before road trials, they have also considered detection and classification in multiple object scenes.

Bats and birds are endangered by wind turbines, leading to a significant amount of fatalities in recent years, mainly due to a direct collision of the animals with the rotor blades or through barotrauma effects. The search for fatalities and the quantification of their number is an important tool to understand the adverse effects of wind turbines on bats and birds. This method has several problems: Cadavers do not always remain on site due to predators, only a part of the relevant area on the ground can be searched manually and not all fatalities can be found. This leads to the fact that searches, although extremely valuable in the analysis of animal mortality, are hardly used in practice due to the high effort and enormous costs. The present work closes this research gap by proposing a radar barrier approach that automatically detects animal fatality in real-time. Therefore, three FMCW radar systems are attached to the tower of a wind turbine at different heights. Subsequent processing of time-of-flight information enables a binary classification of whether a fatality was observed or not. This study presents the results from numerical simulations as well as preliminary experimental results for a proof of concept.

This work deals with the entire process of target detection and ranging by a sky-wave over-the-horizon radar (OTHR) computational model simulation. The different processing stages of the transmitted signal along its space–time trajectory from transmission to digital signal processing are modelled. With this simulation tool a moving target present over the sea can be detected according to a set of given initial conditions together with the ionosphere model inputs and the target electromagnetic model. Initial conditions as well as the modulation and filtering options among other parameters of the model can be set easily. The present work is intended to be a further contribution to OTHR studies, providing a user-friendly tool of easy application in order to improve a radar design, facilitate its implementation, as well as for debugging algorithms and signal processing techniques.

A clutter suppression method based on a combined method of clutter map and K-SVD algorithm has been proposed, which aims at reducing false alarm rate of air surveillance radar caused by wind farm clutter. First, a clutter map, based on the fixed position and Doppler spectral width characteristics of wind turbines, is established using constant false alarm rate detection. The obtained clutter map is then used to process the traces detected by radar. To retain the target while the clutter is suppressed, the traces in the clutter area are processed using the K-SVD algorithm, which is first proposed to mitigate wind turbine clutter in this study. The experimental results of measured data show that the false alarm caused by the wind turbines can be effectively reduced.

Based on the mean arrival time and temporal broadening of pulsed waves propagating through turbulent media, this study investigates the ionospheric effects induced by irregularity on medium-earth-orbit synthetic aperture radar (MEOSAR) range imaging quality. Considering radar signal two-way propagating through ionosphere irregularity, an analysis model for ionospheric effects induced by irregularities on MEOSAR range imaging quality is established based on the system characteristics of MEOSAR and ionospheric irregularity parameters. The mode extends the application of pulsed waves propagating through turbulent media to the field of synthetic aperture radar (SAR), and it can be used to analyse the effects on MEOSAR induced by both background ionosphere and ionospheric irregularity. The degradation of range image quality, including deterioration of resolution and shift in the image, due to the ionospheric irregularities is analysed. It is found that the effects induced by ionospheric irregularities will be very serious when the inner and outer scales of ionospheric irregularities are small or the fluctuation is strong. Furthermore, the degradation of resolution and the shift of image will become more and more serious with the increase of SAR orbit height. The results can help to evaluate ionospheric irregularity effects on range imaging for MEOSAR and provide a foundation for the design of MEOSAR.

The authors consider radar classifications of activities of daily living, which can prove beneficial in fall detection, analysis of daily routines, and discerning physical and cognitive human conditions. They focus on contiguous motion classifications, which follow and commensurate with the human ethogram of possible motion sequences. Contiguous motions can be closely connected with no clear time gap separations. In the proposed approach, they utilise the Radon transform applied to the radar range-map to detect the translation motion, whereas an energy detector is used to provide the onset and offset times of in-place motions, such as sitting down and standing up. It is shown that motion classifications give different results when performed forward and backward in time. The number of classes, thereby classification rates, considered by a classifier, is made varying depending on the current motion state and the possible transitioning activities in and out of the state. Two different examples are given to delineate the performance of the proposed approach under typical sequences of human motions.

This study considers the Bayesian programming methodology for recognition and classification of radio emission sources. A mathematical model of Bayesian programming proposes forming a family of probability distributions based on known parameters contained in a training sample (database). The correlations between the object classes have been estimated according discussed methodology. The received assessment has been used to separate procedures of recognition and classification of radar emission sources. The simulation of methodology has carried out for four parameters of radar signals (frequency range, pulse width, pulse repetition interval and radar rotation frequency) by used database with 346 classes and 16 types of radar. Based on the existing database of radar emission sources, it is possible to predict the probability of class recognition for the general population of objects, if its distribution is known. This study demonstrates the consistency of the Bayesian programming methodology for object identification in ELINT systems.

Time delay estimation (TDE) between receivers plays a significant role in various fields such as radar, sonar, global positioning system, and underwater communications. TDE is the first step in the localisation and tracking of the source. Estimation of the time delay of a passive sonar system in a reverberant environment is one of the most challenging tasks. For some decades, the well-known generalised cross-correlation (GCC) method has been implemented to estimate the time delay. By using various pre-filters in the GCC method, the quality of TDE has improved in various cases. In this work, the use of wavelet denoising for TDE in a passive sonar system has been proposed. The proposed method uses the denoised signal in a maximum likelihood estimator in the reverberant environment to estimate the time delay. The proposed method is efficient for the underwater reverberant environment.

Recognition of multi-function radar (MFR) work mode in an input pulse sequence is a fundamental task to interpret the functions and behaviour of an MFR. There are three major challenges that must be addressed: (i) The received radar pulses stream may contain an unknown number of multiple work mode class segments. (ii) The intra-mode and inter-mode knowledge of a modern MFR may be too flexible and complicated to be represented and learned through traditional hand-crafted features and learning models. (iii) The variable duration of each enclosed work mode makes the identification of the transition boundaries of adjacent modes difficult. To address these challenges and implement automatic recognition of MFR work mode sequences at a pulse-level, this study develops a novel processing framework based on a time series representation of MFR work mode sequence and sequence-to-sequence (seq2seq) long short-term memory network. The proposed method can not only automatically recognise multiple complexes modulated work mode classes in a pulse sequence. Still, it can also accurately identify the transition boundaries between each class by labelling the class information for each pulse. The experimental results showed the extended capabilities and improved performance of the proposed method over the state-of-the-art work mode classification methods.

Most existing work on radar classification of micro-drones assumes that the received signal is wholly reflected from a single micro-drone. However, when there are multiple micro-drones in the observed scene, the superimposition of their micro-Doppler signatures increases the classification difficulty. In particular, it is even more challenging to determine if a specific type of micro-drones exists. In this study, a method for recognition of multiple micro-drones based on their micro-Doppler signatures via dictionary learning is introduced. First, the dictionary is learnt for each type of micro-drone by using the K-SVD algorithm on cadence-velocity diagrams (CVD) of training samples. The CVD is obtained by computing the Fourier transform of the time series of a complex time-frequency spectrogram. Subsequently, the sparse representation of the CVD of multiple micro-drones is obtained by the orthogonal matching pursuit algorithm with the learnt dictionary. Finally, a threshold detector is applied to the sparse solution in order to extract the components of multiple micro-drones. Experimental results using measured data, which are collected from hovering drones by a continuous-wave radar in an indoor environment, show that this dictionary-learning-based method achieves a recognition performance of 93% when half of the measured data are used for training.

Most research regarding passive radar exploiting `illuminators of opportunity', such as FM radio, has focused on detecting and tracking targets. This study explores adding automatic target recognition (ATR) capabilities to such systems. The ATR algorithms described here use the radar cross-section (RCS) of potential targets, collected over a short period of time. The received signal model accounts for aircraft position and orientation, propagation losses, and antenna gain patterns. One proposed algorithm uses a coordinated flight model to estimate aircraft orientations, while a more sophisticated algorithm uses an extended Kalman filter to estimate the target orientations along with measures of uncertainty in those estimates. In both cases, the orientations are estimated using velocity measurements obtained from a tracking algorithm. The radar return of each aircraft in the target library is simulated as though each is executing the same manoeuvre as the target detected by the system. To improve the robustness of the result, the more sophisticated algorithm jointly optimises over feasible orientation profiles and target types via dynamic programming.

Automatic modulation classification of radar signals, which plays a significant role in both civilian and military applications, is researched in this study through a deep learning network. In this study, a novel network combined a shallow convolution neural network (CNN), long short-term memory (LSTM) network and deep neural network (DNN) is proposed to recognise six types of radar signals with different signal-to-noise ratio (SNR) levels from -14 to 20 dB. First, raw signal sequences in the time domain, frequency domain and autocorrelation domain are as input for a shallow CNN. Then the features extracted by CNN will be the input of LSTM network. Finally, DNNs will output the signal modulation types directly. The simulation results demonstrate that the accuracies in autocorrelation domain are all more than 90% at -6 dB and close to 100% when SNR > -2 dB. The recognition performances of the three domains are compared. Compared with other recognition methods, the proposed method has higher average accuracy and better performance under low SNR condition. The measured results show that the proposed method has achieved high accuracies of common four kinds of measured radar signals.

To deal with problems of uncertain modulations and multiple pulse widths in pulse waveforms (PWs) during the identifying procedure, a novel specific emitter identification (SEI) method based on PW images (PWIs) and convolutional neural network is proposed. In the method, a more accurate signal model is built with considering the rising, steady and falling part of the whole PW based on actual radar pulse signals. PWI achieves transforming time-domain waveforms to 2D binary images as an SEI analysis feature. To match the PWI feature, a convolutional neural network with the small convolutional kernel is designed to extract the subtle features and finish the supervised training. By tuning the parameters of the convolutional neural network, it completes a balance of consuming time and identifying accuracy. Simulations and experiments indicate that the proposed method outperforms the existed methods on identifying radar individuals with uncertain modulations and multiple pulse widths in the intercepted pulse signals.

Funding: UK Science and Technology Facilities Council ST/N006569/1 (DR). This study presents a convolutional neural network (CNN) based drone classification method. The primary criterion for a high-fidelity neural network based classification is a real dataset of large size and diversity for training. The first goal of the study was to create a large database of micro-Doppler spectrogram images of in-flight drones and birds. Two separate datasets with the same images have been created, one with RGB images and other with grayscale images. The RGB dataset was used for GoogLeNet architecture-based training. The grayscale dataset was used for training with a series architecture developed during this study. Each dataset was further divided into two categories, one with four classes (drone, bird, clutter and noise) and the other with two classes (drone and non-drone). During training, 20% of the dataset has been used as a validation set. After the completion of training, the models were tested with previously unseen and unlabelled sets of data. The validation and testing accuracy for the developed series network have been found to be 99.6% and 94.4% respectively for four classes and 99.3% and 98.3% respectively for two classes. The GoogLenet based model showed both validation and testing accuracies to be around 99% for all the cases. Postprint

This study examines moving target detection for airborne radar in heterogeneous environments. The non-homogeneity of the clutter may lead to a shortage of training samples in practice. In contrast to detectors assuming rich training data, the authors consider a knowledge-aided treatment for training data-limited scenarios by incorporating a priori knowledge of the clutter subspace. The proposed detector is developed within the generalised likelihood ratio test framework, based on the sparsity of the clutter response in the space-time domain. It does not require training samples and allows the usage of uncertain a priori knowledge of the clutter subspace, which may be erroneous. The maximum likelihood estimate of the target amplitude is obtained via oblique projection, which provides more accurate estimates compared with the conventional orthogonal projection. Moreover, they jointly exploit the sparsity of the clutter signal, the block-Toeplitz nature of the clutter covariance matrix, and the a priori knowledge about the clutter subspace, and develop an atomic norm minimisation approach to achieve super-resolution estimation of the clutter response. The simulation results indicate that the proposed detector outperforms several state-of-the-art methods.

In the introduction, the topic of passive radar jamming and the possibilities of using dipole reflectors (Chaff) to provide radar camouflage are described. This study further describes the two-dimensional and three-dimensional mathematical and physical simulations and rows of Chaff in free space. It then concentrates on the rotation and distances of Chaff with regard to electric field and radar cross-section (RCS) of simulating Chaff puffs in limited space. The theoretical analysis is supplemented by complex computer based simulation. Finally, the required amount of Chaff for protecting aircraft in airspace is determined. This study provides a detailed overview of the physical processes occurring in a limited number of Chaff in open space by the use of a modern simulation software tool.

The study utilises two shape factors as the radar feature set for the recognition of a mid-sized aircraft target in the high-frequency (HF) band. The two shape factors are the dihedral and tilt features of the aircraft parts. The determination of the two shape factors is feasible through the optimum polarisation states of the target's complex natural resonances in the HF band. The identification algorithm applies a resonance-weighted measure to the feature set in a supervised learning approach to discriminate between two targets of the same class that slightly differ in shape. The rate of correct recognition within a sector of the target aspect exceeded 80% for receiver sensitivity above −10 dB.

Accurate antenna motion-measurement is an essential pre-condition for synthetic aperture radar (SAR) to produce well-focused images. The measurement accuracy at high and low frequencies affects the quality of images and the geolocation error. Hence, motion compensation and focusing techniques require a minimum degree of accuracy in the radar motion measurements. The Inertial Measurement Unit (IMU) sensor selection process is done based on the SAR parametric requirements. The IMU instruments have inherent limited precision and accuracy, which are significantly related to the technology and the cost. The navigation algorithm may improve the precision and accuracy to overcome the limitations of the IMU sensor. This study intends to improve motion measurement accuracy using frequency-weighted state estimation. The proposed algorithm shows an error reduction in state estimation and an improvement in some parameter indices which are in relation to SAR image quality and geolocation. Two innovative algorithms are proposed in this study. The first one is the frequency-weighted Kalman filter which is an improved form of the Kalman filter. Another one is multi-resolution decomposition which is an approach to evaluate the SAR motion measurement system.

Feature extraction for radar high-resolution range profiles (HRRPs) is an essential procedure for robust target recognition based on radar HRRPs. It is almost always better to utilise multiple HRRPs from the target than a single HRRP. The commonly used recognition techniques are not robust enough to extract features from HRRP sequences, especially when there is not enough training data. In this study, the authors present a compact model named the infinite conditional restricted Boltzmann machine (iCRBM) for jointly feature learning and recognition of radar HRRPs. The model is developed from the conditional restricted Boltzmann machine, which is a powerful model for representing high-dimensional and multi-modal conditional distributions. Unlike ordinary RBMs, iCRBM can automatically select the model complexity according to the data, which especially suits for the case of large data dimensionality and small training dataset. A multi-scale structure is also proposed for learning multiple scales of temporal/spatial dependencies. Two types of radar HRRP recognition experiments were conducted to validate the efficiency and robustness of iCRBM: the multi-view HRRP recognition and the small aperture HRRP recognition. The results indicate that, the proposed model can be more adaptive at learning the feature of HRRPs than other common methods. The multi-scale structure is helpful for better recovering missing HRRP samples from non-consecutive sequences.

In this study, the authors propose an energy-efficient adaptive transmit beamspace design for cognitive frequency diverse array (FDA) radar to efficiently detect and track the target in a cognitive way, particularly, for the case when the target is lost in one scan. Additionally, an algorithm is presented for a selection of the illumination range-angle sector to recover a lost target. Moreover, an algorithm is proposed to discriminate the target from interferers, while an FDA beampattern time-variance that is normally ignored in the literature is also considered. Finally, the transmit beamspace is adaptively updated to illuminate the predicted range-angle-based area of interest in each cognitive scanning cycle. The proposed design generates a single maximum beam pattern by using a set of non-uniform frequency offsets along an FDA. Numerical results show that the proposed method achieves improved target detection and tracking performance as compared to the standard frequency diverse and phased-array radars due to its increased output signal-to-interference-plus-noise ratio, while improved energy efficiency is achieved due to an adaptive range-angle sector selection algorithm.

Jet engine modulation (JEM) are micro-Doppler features obtained from the radar data of aircraft, including jet aircraft and helicopters, and generally viewed as unique phenomenon suitable for identifying aerial targets. The authors demonstrate however, that JEM are not unique, and that the rotating wheel spokes on the wheels of ground vehicles produce a modulation similar to the rotating rotor blades of jet engines and helicopter rotor blades. They also develop a mathematical model to analyse the radar signal scattering mechanism of wheel spokes, and also employ a Ku-band Doppler radar system to collect radar echoes from a wheeled vehicle and a helicopter. Both the theoretical and experimental results verify the existence of rotor blade modulation (RBM) phenomena from helicopters and wheel spoke modulation (WSM) phenomena from vehicles. Both RBM and WSM are similar to JEM features. This fact indicates that JEM is a necessary but insufficient condition for the identification of aerial targets, and additional recognition features inherent in surface surveillance radar are needed to distinguish the JEM spectra arising from aerial targets and ground vehicles.

The azimuthal imaging can be acquired by the orbital angular momentum (OAM) wave. In addition, a large-scale OAM mode number is necessarily required for high azimuthal resolution. To reduce the required OAM mode number, an azimuthal imaging method based on compressed sensing and OAM mode selection is presented. To be specific, the system architecture and the mathematical model of the OAM imaging radar are established. Meanwhile, an efficient OAM generation method with rotational antenna is proposed. Moreover, experiments show that different radar cross sections (RCSs) of a specific complex target can be obtained by different OAM waves, which are generated by the proposed method. It is a delight to find that the signal-to-noise ratio gain can be achieved due to the appropriate OAM mode selection in the experiment. Since different OAM waves lead to different RCSs of complex target, the OAM modes corresponding to high RCSs are selected to measure the azimuth for low complexity. Furthermore, simulations are conducted to evaluate the performance of the proposed azimuthal imaging method for the general complex target. Using Monte Carlo simulations, it is found that the accuracy of azimuthal imaging can be guaranteed by the proposed azimuthal imaging method with only approximately 6% selected OAM modes.

Ground moving target indicator (GMTI) RADAR system frequently incorporates space-time adaptive processing for successful moving targets discrimination in a highly cluttered environment. The minimum detectable velocity (MDV) of the target gets affected due to the added clutter impairments such as foliage penetration and angle independent channel mismatch. A novel clutter model is postulated in this work that includes assessment enacted in view of both angle and Doppler clutter impairments at a time instead of their individual analysis as in the reported literature. The performance of the GMTI RADAR in terms of the signal-to-interference-plus-noise ratio loss is evaluated using the extrusive beamforming methods namely sample matrix inversion and principal component canceller. In high decorrelated clutter environment, the later outperforms in terms of the MDV, making use of a lesser number of training snapshots. The approach is validated with mathematical modelling and numerical simulations.

The deployment of deep neural networks in real-world radar-based human activity classification is largely hindered by both the high computational cost and the large amount of training samples. In this study, the authors propose a method to simultaneously reduce the computational burden and the number of labelled training samples. Different from previous transfer learning methods that simply prune fully-connected layers and modify the weights of the convolutional layers, they enforce filter-level sparsity in the transfer learning from ImageNet to the micro-Doppler measurements. Through the sparsity-driven transfer learning, unimportant convolutional filters can be identified and then be pruned. Therefore, a light but effective transfer learned net can be obtained. The experiments demonstrate the sparsity-driven transfer learned VGG-19 Net not only outperforms convolutional neural networks trained from scratch by nearly 10% accuracy but also gives an 11 × reduction in the number of parameters and a 10 × reduction in computing operations compared with the original VGG-19 Net.

Autonomous navigation of platforms in complex environments has a key role in many applications. However, environmental conditions could negatively affect the performance of electro-optical sensors. Hence, the idea of using radar odometry (RO) has been recently developed. However, it suffers from the presence of outliers in the scene as its electro-optical counterparts. This work presents a method to classify radar echoes as inliers or outliers for two-dimensional RO, based on their range rate and bearing angle. The range rate and bearing angle are, in fact, combined to give a classification value, different for each target. At each acquisition time, the median of all classification values is computed. Since classification values of stationary targets, i.e. the inliers, cluster around the median, while moving targets, i.e. the outliers, exhibit larger distance from the median, stationary targets and moving targets can be separated. This is also useful for sense-and-avoid purposes. The method has been tested in the simulated scenario to show effectiveness in detecting outliers and in a real-case scenario to demonstrate significant improvement in the reconstruction of the trajectory of the platform, keeping the final error around 10% of the travelled distance. Further improvement is envisaged by integrating the method in the target tracking strategy.

RAD-DAR is a frequency-modulated-continuous-wave radar demonstrator, working at X-band, completely designed and constructed by the authors' research group following the ubiquitous-radar concept. After introducing its main hardware and software blocks, with special emphasis to its off-line signal processing and data processing, this study presents the results of a drone detection test, where RAD-DAR achieved a DJI-Phantom 4 detection and tracking at a range up to 3 km. The system performance is discussed with different flights including an attack manoeuvre, and a free flight, which helps to highlight their advantages in surveillance tasks due to the RAD-DAR staring nature. Furthermore, range, speed and azimuth accuracies are discussed, considering the drone global positioning system data. Finally, this study shows a statistical radar cross-section study based on the processed data of the drone and attempts to classify this target by means of Swerling models.

Radar systems emit a time-varying signal and measure the response of a radar-reflecting surface. In the case of narrowband, monostatic radar signal domain, all spatial information is projected into a radar cross-section (RCS) scalar. The authors address the challenging problem of determining shape class using monostatic RCS estimates collected as a time series from a rotating object tumbling with unknown motion parameters under detectability limitations and signal noise. Previous shape classification methods have relied on image-like synthetic aperture radar or multistatic (multiview) radar configurations with known geometry. Convolutional neural networks (CNNs) have revolutionised learning tasks in the computer vision domain by leveraging images and video rich with high-resolution two-dimensional (2D) or 3D spatial information. They show that a feed-forward CNN can be trained to successfully classify object shape using only noisy monostatic RCS signals with unknown motion. They construct datasets containing over 100,000 simulated RCS signals belonging to different shape classes. They introduce deep neural network architectures that produce 2% classification error on testing data. They also introduce a refinement network that transforms simulated signals to appear more realistic and improve training utility. The results are a pioneering step toward the recognition of more complex targets using narrowband, monostatic radar.

As of the large receiving gain from radar antenna and the little difference with targets in the time-frequency-spatial domain, the main-lobe interference has become an important challenge to radar detection. In recent years, suppressing main-lobe interference by using polarisation filtering has gained significant theoretical and practical progress. However, when there are multiple interference sources occurring in the radar main lobe, the performance of all existing methods is badly limited. With the idea of feature identification, a novel target detection method based on polarisation-space joint spectrum characteristic is developed, which effectively utilises the polarisation scattering feature difference between the target returns and the interference signal, instead of trying to suppress interference energy as usual. This proposed method can effectively enhance the capability of radar target detection under the background of multi-point source interference in main lobe. The simulation results show that the proposed method can detect the target in the background of the main-lobe multi-point source interference.

This study proposes an artificial neural network-based method to classify road structures for automotive radar systems. Generally, road structures generate unwanted echoes called clutter, which degrades the performance of target detection, and may cause critical errors in autonomous driving modes. However, recognition and classification of each individual structure type is very difficult, because there are various types of structures made by different materials which cause ambiguity in the recognition and classification process. To deal with these classification ambiguities, the authors propose an artificial neural network-based recognition and classification method. Road structures are classified by applying artificial neural network to the frequency domain-received signals of automotive radar systems. Once the neural network has been trained with the received signals, it is used to determine the type of road structure with instantaneous received signals. The classification performance was evaluated by experimenting with the number of antenna elements and radar snapshots. In addition, to find the most suitable artificial neural network structure, the authors experimented with the number of nodes and layers. After completing a period of deep learning using actual experimental data, the total classification accuracy was about 95%.

In this study, to reduce the performance degradation due to iron-clutter, the authors propose a suppression method for stationary clutter under iron-structures for an automotive radar using a frequency-modulation continuous-wave (FMCW) radar. The presence of a large number of iron-structures such as iron-bridges, soundproof walls, and iron-tunnels creates a harsh environment that deteriorates the associated radar detection performance owing to clutter. To mitigate the adverse effects of clutter, they propose a clutter suppression scheme using up- and down-chirp signals based on an analysis of signal characteristics. In particular, it is shown that the stationary clutters can be suppressed by making that the phases of the up-chirp and down-chirp are similar. To experimentally verify the proposed method, 77 GHz FMCW radar data for adaptive cruise control was acquired along a road on which several iron-structures were located. The experimental results show that using the proposed clutter suppression scheme, the frequency of the moving target was detected well using the proposed clutter suppression scheme in an iron-clutter-rich environment. Through the application of the proposed clutter suppression scheme, stable target detection performance of the radar can be obtained.

With the rapid development of the integrated circuit manufacturing technology and the radar technology, the demand for miniaturisation, low power consumption, and real-time performance of synthetic aperture radar (SAR) imaging system is becoming higher and higher. By designing reconfigurable IP cores and configuring parameters through CPU, a reconfigurable missile-borne SAR imaging SoC chip based on IP reuse is designed to meet this urgent demand. Using 0.13μm CMOS process, the SoC chip has been taped out and verified successfully. Compared with the traditional ‘DSP + FPGA’ platform, this chip can ensure the real-time performance of the missile-borne SAR imaging system and has more advantages in scale and power consumption.

In part II of the passive inverse synthetic radar (P-ISAR) sequence, implications of narrow signal bandwidths on image formation procedures are considered within the previously developed P-ISAR framework of part I. Specifically, the case of single-channel digital television imaging systems is evaluated. For average sized airborne targets illuminated with narrow signal bandwidths, the response is confined to one or two down-range resolution cells, which has significant impact on both inverse synthetic aperture radar (ISAR) resolutions and motion compensation procedures. Simulated results demonstrate that a non-parametric autofocus method, phase gradient algorithm, does not necessarily converge to a properly focused image. Contrast and entropy metrics are also shown to be insufficient for the narrowband imaging case. A parametric Doppler centroid approach is proposed for phase compensation, with resulting imagery closely resembling clairvoyant reconstruction. Finally, the proposed architecture is evaluated in full-scale experiments of commercial air targets, demonstrating the potential for discriminating large and small aircrafts.

To monitor the sea-state and detect low-speed targets beyond the line-of-sight limitation, high-frequency (HF) sky-wave radar needs to extract the sea-clutter Doppler spectrum and suppress radio-frequency interference (RFI). However, this is hard for traditional radar processing techniques in practice. Instead, this study proposes to tackle these problems based on image recognition methods. By converting the range-Doppler (RD) map into a grey image, the extraction of sea-clutter and RFI regions becomes two problems of image segmentation. The sea-clutter region is extracted by the distance regularised level set evolution algorithm, based on a binary image which uses threshold segmentation to roughly classify sea-clutter pixels and noise pixels. To extract the RFI region, the authors develop a straight line detection algorithm which employs the LOG operator to mitigate the sea-clutter and highlight the RFI lines. Simulation results demonstrate that the proposed approaches are effective in the RD map with undesired RFI lines near the sea-clutter.

In this study, a scale mixture of normal distributions model is developed for classification and clustering of radar emitters. A radar signal is characterised by a pulse-to-pulse modulation pattern and is often partially observed. The proposed model can classify and cluster different radar emitters even in the presence of outliers and missing values. The classification method, based on a mixture model, focuses on the introduction of latent variables that give us the possibility to handle sensitivity of the model to outliers and to allow a less restrictive modelling of missing data. A Bayesian treatment is adopted for model learning, supervised classification and clustering. The inference is processed through a variation Bayesian approximation. Some numerical experiments on realistic data show that the proposed method provides more accurate results than state-of-the-art classification algorithms.

Through-the-wall radar (TWR) with ultra-wideband (UWB) signals is a promising technology for three-dimensional monitoring systems used for rescue and security applications. To develop a suitable imaging algorithm to deal with UWB radar data, the range points migration (RPM) method was enhanced for the TWR observation model. However, there are some problems in the actual application because multiple reflection signals from the walls or targets produce a ghost image. This research extends the RPM-based TWR imaging method to convert a ghost image to an accurate one by exploiting the double-bounced signals discriminated by the Doppler velocity. Such discrimination can be achieved by prior estimations for the motion and location of the target provided by the Doppler-associated RPM algorithm. The finite-difference time-domain-based numerical simulation, with the assumption of multiple surrounding walls, demonstrated that the authors proposed method produced an accurate object by converting a ghost image to an actual image.

The authors present cognitive automotive radar (CARr) as a significant capability to the sensing technology suite of autonomous driving vehicles. CARr is a closed-loop intelligent radar system utilising the automotive frequency bands of 24-25 and 76-77 GHz for the purposes of ground vehicle target signature identification. The authors consider specific cases of ground vehicle recognition and ground vehicle class identification in the presence of aspect angle uncertainty using forward-looking automotive radar. The transmit-adaptive waveforms are based from signal-to-noise ratio and mutual information optimisation metrics. In this study, two new adaptive waveform techniques with the flexibility to accommodate angular uncertainty probability distributions are introduced. The classification performance of these new waveforms is compared against the receive-adaptive wideband pulsed and other transmit-adaptive waveforms in the presence of angular uncertainties characterised by the uniform and truncated normal distributions. To ensure the validity of results, high-fidelity electromagnetic simulated radar cross-section signatures generated from scaled-to-true-physical-size ground vehicle computer-aided design models are utilised.

Driven by the demand to develop self-driving cars, automotive radars evolve rapidly at this time. Current high end radars therefore have very little in common with the radars that were developed just a few years ago and are being built into today's car series. Anyway, since at least two decades such sensors are referred to as imaging radars, although the set of possible applications has changed significantly. This study displays the leap in the evolution of automotive imaging radars on the example of the Astyx HiRes radar and elucidates the additional information that becomes available for autonomous driving functions like freespace modelling and target classification. In particular, it will be shown that the image quality of a modern high end radar suffices to apply computer vision techniques that were reserved for optical or light detection and ranging (LIDAR) images until now.

Multibeam radar (MBR) systems based on waveform diversity require a set of orthogonal waveforms in order to generate multiple channels in transmission and extract them efficiently at the receiver with digital signal processing. Linear frequency modulated (LFM) signals are extensively used in radar systems due to their pulse compression properties, Doppler tolerance, and ease of generation. Here, the authors investigate the level of isolation between MBR channels based on LFM chirps with rectangular and Gaussian amplitude envelopes. The orthogonal properties and the mathematical expressions of the isolation are derived as a function of the chirp design diversity, and specifically for diverse frequency slopes and frequency offsets. The analytical expressions are validated with a set of simulations as well as with experiments at C-band using a rotating target.

In this study, two radar sensors with angular diversity are used to recognise dynamic hand gestures by analysing the sparse micro-Doppler signatures. The radar echoes are firstly mapped into the time–frequency domain through the Gaussian-windowed Fourier dictionary at each radar sensor. Then the sparse time–frequency features are extracted via the orthogonal matching pursuit algorithm. Finally, the sparse time–frequency features at two radar sensors are fused and input into the modified-Hausdorff-distance-based nearest neighbour classifier to achieve the dynamic hand gesture recognition. The experimental results based on the measured data under three different experimental scenes demonstrate that (i) the recognition accuracy can be improved by fusing the features extracted at two radar sensors when each radar sensor works well on its own; (ii) the recognition accuracy produced by feature fusion keeps satisfactory even if one of the radar sensors has poor performance, which means that the feature fusion can improve the robustness of the recognition system; and (iii) it would be more helpful if the line-of-sights of the two radar sensors are set to be orthogonal to each other.

Radar emitter classification (REC) is an essential part of electronic warfare (EW) systems. In REC tasks, after deinterleaving, the intercepted radar signals are classified into specific radar types. With new radar types arising and the electromagnetism environment getting complicated, REC has become a big data problem. Meanwhile, there exist inconsistent features among samples. These two problems can affect the performance of classification. In this work, first, the authors designed a novel encoding method to deal with the inconsistent features. High-dimension sequences of equal length are generated as new features. Then a deep learning model named unidimensional convolutional neural network (U-CNN) is proposed to classify the encoded high-dimension sequences with big data. A large and complex radar emitter's dataset is used to evaluate the performance of the U-CNN model with the encoding method. Experiments show that the authors' proposal gains an improvement of 2-3% in accuracy compared with the state-of-the-art methods, while the time consumed for identifying 45,509 emitters is only 1.95 s using a GPU. Specifically, for 12 indistinguishable radars, the classification accuracy is improved about 15%.

As tracked vehicles play significant roles in a battlefield, effective jamming measures are necessary to protect them from being perceived by a hostile radar. Moving tracked vehicles usually exhibit strong Doppler and micro-Doppler signatures. Therefore, the jamming signal should include micro-Doppler modulation generated by metallic caterpillars for successful deceptive jamming. Based on detailed analysis of kinetic characteristics of tracked vehicles, this study proposes a novel deceptive jamming method for tracked vehicles against a continuous-wave ground surveillance radar. To guarantee the fidelity of the deceptive jamming, this method performs both translational modulation for rigid parts and micro-Doppler modulation for the caterpillars. Moreover, the translational modulation function is generated partly off-line to improve the computational efficiency. To evaluate the performance of the proposed approach quantitatively, evaluation indices, such as human visual system and wavelets weighted mean square error, are designed. Simulation results are presented to verify the validity of the proposed method.

The latest works on the radar pulse de-interleaving in electronic support measures (ESM) systems have not completely solved the missing pulses problem yet. The authors present a pulse detection algorithm that aims at, if not eliminating it, diminishing the rate of pulses not detected because of the fact of being superimposed in other ones. Experiments set up on an ESM system based on software-defined radio (SDR) showed that the algorithm detection rate was near a hundred per cent and the false alarm rate was near zero when pulse amplitudes were >3.9 mV in the SDR input. This work is suggested for the areas of digital and radar signal processing and electronic warfare.