Focus-before-detect for radar target detection

The multivariate range function of the high-speed maneuvering target induces modulations on both the envelop and phase, i.e., the range cell migration (RCM) and Doppler frequency migration (DFM) which degrade the long-time coherent integration used for detection and localization. To solve this problem, many long-time coherent integration methods have been proposed. Based on mechanisms of typical methods, this paper names two signal processing modes, i.e., processing unification (PU) mode and processing separation (PS) mode, and presents their general forms. Thereafter, based on the principle of the PS mode, a novel long-time coherent integration method, known as the generalized dechirp-keystone transform (GDKT), is proposed for radar high-speed maneuvering target detection and localization. The computational cost, energy integration, peak-to-sidelobe level (PSL), resolution, and anti-noise performance of the GDKT are analyzed and compared with those of the maximum likelihood estimation (MLE) method and keystone transform-dechirp (KTD) method. With mathematical analyses and numerical simulations, we validate two main superiorities of the GDKT, including (1) the statistically optimal anti-noise performance, and (2) the low computational cost. The real radar data is also used to validate the GDKT. It is worthwhile noting that, based on closed analytical formulae of the MLE method, KTD method, and GDKT, several doubts in radar high-speed maneuvering target detection and localization are mathematically interpreted, such as the blind speed sidelobe (BSSL) and the relationship between the PU and PS modes.

Deep neural network-aided coherent integration method for maneuvering target detection

Radar fast long-time coherent integration via TR-SKT and robust sparse FRFT

This study proposes the Space–time Radon–Fourier transform (STRFT) for wideband digital array radar (DAR). It is a novel coherent detection technique that jointly realise wideband digital beamforming (DBF), range compression and long-time coherent integration. The likelihood ratio detector (LRT) is first derived based on the three-dimensional (3D) signal model of moving targets, which is the output of the STRFT on the 3D range-compressed echoes. Some transform properties of the proposed STRFT are also discussed. This include the 3D impulse response, 3D translational invariance, multi-target linear additivity, linear SNR gain in additive white Gaussian noise (AWGN), as well as the 3D correlation function of transformed AWGN. Also, the fast implementation of STRFT is proposed in the element-pulse-range frequency domain, which is realised via the two dimensional Chirp-Z transform (CZT) based on fast Fourier transform (FFT) operators. The proposed method may remarkably reduce the computational burden for modern wideband DAR. Finally, some numerical experiments are provided to demonstrate the effectiveness of the proposed methods.

Coherent integration is a conventional approach to improve detection ability in radar applications, whereas its performance is decreased by migration through range cell due to the motion of target, limiting the time of integration. This problem becomes more severe in terahertz band. This paper introduces Keystone transform to terahertz radar in order to realize long-time coherent integration, aiming to get a successful detection of targets. Moreover, doppler ambiguity is resolved based on intra-pulse interference processing. A simulation is conducted to demonstrate the effective improvement for target detection using Keystone transform in terahertz radar.

High-speed, low-observable, and maneuvering target detection is always the difficult problem for the early warning and detection of space target, ballistic target, and passive radar signal processing. The high-order motion information of maneuvering target is employed in this paper, such as acceleration and jerk. Accordingly, several long-time coherent integration methods are demonstrated and illustrated, i.e., Radon-fractional Fourier transform (RFRFT), Radon fractional ambiguity function (RFRAF), Radon-linear canonical ambiguity function (RLCAF), and phase differentiation-Radon-Lv's distribution (PD-RLVD). These methods can compensate the across range unit (ARU) and Doppler frequency migration (DFM) effects simultaneously during the long-time integration. The background clutter and noise can be well suppressed with higher integration gain. Experimental results using real radar data show the superiority of the proposed methods.

Signal integration, as an effective method of detecting weak target, is widely used in areas of radar, sonar, etc. In previous studies of long-time coherent integration, researchers usually established a multi-pulse echo model with linear frequency modulation (LFM) signal due to its good compression performance and large Doppler tolerance. Then, perfect analytical formula can be deduced to compensate for range migration and Doppler spread, which is helpful in analyzing the mechanism of long-time coherent integration in depth. However, besides LFM, a wide variety of signal waveforms are also used in modern sonar and underwater guidance system to meet the requirements for diverse applications. For instance, continuous wave (CW) pulse is often used in signal detection, high resolution direction of arrival (DOA) estimation, and velocity estimation, while large time-bandwidth product waveforms such as modulated signal, coded signal, and pseudo-random signal are utilized for special tasks like anti-interference detection, channel matching, and active concealed detection. Therefore, the formulas and corresponding instructive conclusions deduced by LFM have no generality when other sonar waveforms are used in pulse integration. In this paper, we focus on long-time coherent integration for arbitrary signal reflected by underwater target moving with a uniform velocity and propose a motion-compensated coherent integration method for arbitrary complex envelop signal. A kind of general ambiguity function (GAF) for transmitted signal is defined to present a unified expression based on GAF for the output of the matched filter. The operation not only helps us to describe and calculate the pulse compression form of the arbitrary complex envelop by using a general mathematical model, but also provides information about the range migration and Doppler frequency shift of the multi-pulse echo, which is needed in pulse range alignment and FFT integration. For the matched filter output expressed by the GAF, Keystone transform is utilized to correct the complex envelop of the multi-pulse echo and eliminate the range walk. Then, Doppler frequency shift is compensated for by performing FFT transform, and the long-time coherent integration for arbitrary complex envelop is realized. To verify the correctness of the proposed method, we carry out the computer simulation on both signal integration and detection performance by using four sonar waveforms, i.e., CW signal, LFM signal, m-sequence phase-coded signal, and Costas frequency hop coded signal. The simulation results show that the proposed motion-compensated coherent integration method is applicable to arbitrary complex envelop signal. We also design an anechoic water tank experiment scheme which can successfully obtain the multi-pulse echoes of constant moving target. The motion-compensated coherent integration of the experimental data of the above-mentioned four waveforms further validates the effectiveness of the proposed method.

At present, after radar target detection, radar target recognition mainly depends on manual judgment. Manual identification relies too much on the operator's personal experience and subjective consciousness, which takes a long time and has a large error. To solve this problem, a radar target recognition method based on deep learning is proposed. By analyzing the differential characteristics of radar echoes of different targets, taking the range profile of radar echo sequence as the data set, a deep residual network classifier model is designed to classify and identify the types of radar targets. In order to improve the generalization ability of the network and avoid the over-fitting problem, the deep convolutional generative adversarial network is used to expand the range profile data set of radar echo sequence. In order to ensure the quality of the generated samples during data set expansion, the expanded samples are screened with the peak signal-to-noise ratio as the index. The experimental results of radar measured data show that this method has a good effect on radar target recognition.

Long-time coherent integration can remarkably improve the detection and motion parameter estimation ability of radar for maneuvering targets. Unfortunately, the linear range migration, quadratic range migration (QRM), and Doppler frequency migration (DFM) within the coherent processing interval seriously degrade the target detection and parameter estimation performance. In this regard, a novel coherent integration approach based on frequency autocorrelation function (FAF) and Lv’s distribution (LVD), i.e., FAF–LVD, is proposed in this paper. The QRM is firstly removed by the second-order keystone transform. Then, the FAF with a variable delay is introduced to remove the DFM, followed by the scaled Fourier transform to achieve coherent integration. Finally, the acceleration and velocity are estimated simultaneously by LVD. Analysis demonstrates that FAF–LVD could be efficiently implemented via complex multiplications, the fast Fourier transform (FFT) and inverse FFT. Detailed comparisons with several representative methods indicate that FAF–LVD achieves a better balance among computational complexity, detection ability, and parameter estimation performance. Extensive simulations are presented to validate the effectiveness of FAF–LVD method.

Traditional radar target detection algorithms are mostly based on statistical theory. They have weak generalization capabilities for complex sea clutter environments and diverse target characteristics, and their detection performance would significantly be reduced. In this paper, the range-azimuth information obtained by scanning radar is converted into PPI image, and a Radar-YOLONet is proposed and used for radar image target detection. The model includes CSPDarknet53, SPP, PANet and Power Non-Maximum Suppression layers. The prediction frame coordinates, target category and corresponding confidence are directly given through the feature extraction network. The network structure strengthens the receptive field and attention distribution structure, and further improves the efficiency of network training. The verification using the constructed X-band navigation radar PPI image dataset shows that compared with the traditional constant false alarm rate detector and the two-stage Faster-RCNN algorithm, the proposed method has significantly improved detection performance under certain false alarm probability conditions. It is more suitable for various environment and target characteristics.

Based on constant acceleration radial distance model, a novel approach to long-time coherent integration is provided. This technique focuses on detecting high speed targets with ARU walk, by compensating the phase fluctuations of target echoes after pulse compression along slow time dimension. The numerical experiments are provided to demonstrate the effectiveness of the proposed method.

High-dimensional feature extraction of sea clutter and target signal for intelligent maritime monitoring network

Long-time coherent integration using frequency hopping signals is a challenging problem for passive bistatic radar due to its frequency hopping characteristics. Apart from range walk, range curve, and Doppler frequency migration, Doppler diffusion caused by frequency hopping characteristics occurs within the observation time, which also lowers the detection performance. To deal with this problem, a novel coherent integration method for frequency hopping signals based on passive bistatic radar is proposed in this paper. In this novel method, range curve and range walk are eliminated by applying generalized Keystone transform. Then, Doppler frequency migration caused by the target's acceleration is compensated for by a parameter search with a designed search scope. Finally, Doppler frequency migration caused by frequency hopping characteristics is compensated for by designing a new acceleration compensation function and a revised rotation factor for Fourier transform. Since migration effects caused by frequency hopping characteristics are considered and compensated for when using frequency hopping signals, the weak target echo can be better integrated in the observation time compared to when using the existing methods. The simulation results and performance analysis illustrate the effectiveness of the proposed method.

With the increasing of pulse duration and bandwidth of the radar signal, the required computational burden of pulse compression processing data becomes heavier. Therefore, in order to meet the requirements of real-time radar target detection, the large-point real-time pulse compression technology research has become an urgent demand for the radar development. Aiming at the large-point echo data of Non-cooperative Bistatic Radar, we propose a real-time pulse compression algorithm which can be implemented on the multicore DSP TMS320C6678 platform. The algorithm is based on large-point FFT algorithm and takes advantage of high-performance floating-point processing capabilities and multi-core peculiarity of TMS320C6678. In this paper, the accuracy and real time of algorithm implementation is analyzed between multicore DSP TMS320C6678 platform and MATLAB platform. The experimental results show that this method has high accuracy, less time-consuming and meet the requirements of radar real-time target detection.

In recent years, the micro-Doppler effect has been proposed as a new technique for signature analysis and extraction of radar targets. The ballistic missile is known as a typical radar target and has been paid many attentions for the complexities of its motions in current researches. The trajectory of a ballistic missile can be generally divided into three stages: boost phase, midcourse phase and terminal phase. The midcourse phase is the most important phase for radar target recognition and interception. In this stage, the warhead forms a typical micro-motion called micro-nutation which consists of three basic micro-motions: spinning, coning and wiggle. This paper addresses the issue of signature analysis of ballistic missile warhead in terahertz band via discussing the micro-Doppler effect. We establish a simplified model (cone-shaped) for the missile warhead followed by the micro-motion models including of spinning, coning and wiggle. Based on the basic formulas of these typical micro-motions, we first derive the theoretical formula of micro-nutation which is the main micro-motion of the missile warhead. Then, we calculate the micro-Doppler frequency in both X band and terahertz band via these micro-Doppler formulas. The simulations are given to show the superiority of our proposed method for the recognition and detection of radar micro targets in terahertz band.