Improving performance in pulse radar detection using neuralnetworks

A new approach using a multilayered feed forward neural network for pulse compression is presented. The 13 element Barker code was used as the signal code. In training this network, the extended Kalman filtering (EKF)-based learning algorithm which has faster convergence speed than the conventional backpropagation (BP) algorithm was used. This approach has yielded output peak signal to sidelobe ratios which are much superior to those obtained with the BP algorithm. Further, for use of this neural network for real time processing, parallel implementation of the EKF-based learning algorithm is indispensable. Therefore, parallel implementation has also been developed     

A novel target detection approach based on adaptive radar waveform design

To resolve problems of complicated clutter, fast-varying scenes, and low signal-clutterratio (SCR) in application of target detection on sea for space-based radar (SBR), a target detection approach based on adaptive waveform design is proposed in this paper. Firstly, complicated sea clutter is modeled as compound Gaussian process, and a target is modeled as some scatterers with Gaussian reflectivity. Secondly, every dwell duration of radar is divided into several sub-dwells. Regular linear frequency modulated pulses are transmitted at Sub-dwell 1, and the received signal at this sub-dwell is used to estimate clutter covariance matrices and pre-detection. Estimated matrices are updated at every following sub-dwell by multiple particle filtering to cope with fast-varying clutter scenes of SBR. Furthermore, waveform of every following sub-dwell is designed adaptively according to mean square optimization technique. Finally, principal component analysis and generalized likelihood ratio test is used for mitigation of colored interference and property of constant false alarm rate, respectively. Simulation results show that, considering configuration of SBR and condition of complicated clutter, 9 dB is reduced for SCR which reliable detection requires by this target detection approach. Therefore, the work in this paper can markedly improve radar detection performance for weak targets.     

基于小波神经网络的PCBA检测

为了实现对PCBA的快速准确检测,提出了基于小波神经网络的检测模型,分析了网络的拓扑结构.仿真显示,该检测模型的误差小,精确度高,检测速度快,提高了SMT生产线上的一次通过率的可靠性.     

Noncoherent approach to through-the-wall radar localization

A noncoherent through-the-wall radar system approach, based on stepped-frequency signal synthesis and trilateration technique, is presented. This approach involves multiple independent monostatic radar units and as such, provides flexibility in positioning the units with various standoff distances and inter-element spacing. The performance of the proposed noncoherent localization system was demonstrated using simulated and real data. The results show that the radar is able to detect and locate multiple targets behind walls     

Intrepid digital microwave. A new approach to bistatic radar

Southwest Microwave, Inc. (SMI) introduced the world's first commercially viable bistatic microwave intrusion detection sensor in 1971. Bistatic microwave has become the paradigm for high security perimeters, and SMI products have become industry standards. With the introduction of the Intrepid Digital Microwave, SMI brings the latest in Digital Signal Processing (DSP) and power and data networking to this proven sensor technology     

Missing radar pulse clutter processing

The degradation of conventional clutter processing filters due to missing pulses resulting from RF interference blanking, ambiguous range returns when no filter pulses are used, or eclipsing is described. The effects of a missing pulse on uncompensated and optimum clutter filters are investigated. The maximum improvement factor method and other methods are compared to the optimum filtering. New methods and results using a minimax log energy search method are described and are shown to provide performance and/or implementation complexity advantages     

A novel pulse TOA analysis technique for radar identification

A new processing technique is presented for identifying individual pulse trains in a buffer load of pulse time-of-arrival (TOA) data from radar warning receiver sets currently in use for electronic warfare (EW). The method is based on the numerical characteristics of the matrix of the differences of the TOA data points. In particular, for a uniform pulse repetition interval (PRI) emitter, this matrix is a symmetrical harmonic Toeplitz matrix the inverse of which has banded tridiagonal form. This form leads to an estimator for the PRI in terms of the trace of the inverted matrix     

A sequential detection approach to target tracking

Sequential hypothesis testing is investigated for multiframe detection and tracking of low-observable maneuvering point-source targets in a digital image sequence. The proposed multiple multistage hypothesis test tracking (MMHTT) algorithm extends tracks formed from sequentially detected target trajectory segments using a multiple hypothesis tracking strategy. The MMHTT algorithm does not require a probabilistic larger maneuver model. Computational efficiency is achieved by using a truncated sequential probability ratio test (SPRT) to prune a dense tree of candidate target trajectories and score the detected trajectory segments. An analytical performance evaluation is presented and confirmed by experimental results from an optical satellite tracking application     

BP神经网络制导弹道导弹的诸元数据制备方法

采用BP神经网络逼近算法,推导、建立了预定关机点及落点坐标与需要速度间的映射关系,并装订上弹,使关机点附近对应的每一个位置都能够映射出相应的需要速度矢量,利用闭路制导关机及导引方法对导弹实施控制。提出了基于BP神经网络制导的数据制备方法,将大量的弹道数据改为以神经网络形式进行制备,有效地缩小了数据存储空间及弹载计算机的计算时间。仿真结果表明,基于BP神经网路的制导方法能够大大提高导弹的制导精度,相应的诸元数据制备方法能够准确地实现神经网络在弹上的映射功能,为该制导方法在弹上的应用奠定了基础。     

A model for target detection with over-the-horizon radar

The effects of including Faraday rotation and multipath on the probability of detecting low-flying, distant, fluctuating and nonfluctuating targets immersed in Rayleigh noise plus clutter are studied. The effect of ionospheric fluctuations is also considered. It is found that both multipath and Faraday rotation strongly influence the detection statistics, with the effect being greatest for linearly polarized targets and less marked for symmetric targets     

A morphological approach to automatic mine detection problems

We consider the problem of detecting mines and minelike targets imaged by multispectral sensors. We propose an algorithm, based on mathematical morphology (MM), that yields accurate detection results in moderately cluttered environments. For targets in heavily cluttered environments, a preprocessing step is employed, based on the maximum noise fraction (MNF) transform, in order to reduce the effect of clutter and enhance the presence of targets. The algorithm is simple, performs well, and requires only approximate knowledge of target size     

Ambiguity function of quadriphase coded radar pulse

Taylor's quadriphase coding (J.W. Taylor, Jr. and H.J. Blinchikoff, ibid., vol.23, no.2, p.156-70, Mar. 1988) is investigated for nonzero Doppler shifts. While the zero-Doppler cut of the ambiguity function (i.e. the autocorrelation) strongly resembles that of the corresponding biphase code, the remaining ambiguity function differs considerably. The ambiguity function of quadriphase code 13 is typified by a diagonal ridge as found in linear FM signals. The ambiguity function of quadriphase code 28A resembles the three parallel ridges of Frank code 16     

Improving intrusion detection radar

The first monostatic microwave intrusion detection sensor with range cutoff was introduced in 1984. This range cutoff circuit as used in the Model 375 and 385 has proven very effective in preventing nuisance alarms beyond a user-defined range. The Intrepid Digital Transceiver introduced in this paper builds upon this proven technology with the addition of a unique digital signal processing routine that measures range to the intruder. Intrepid Digital Transceiver alternately transmits pulses at two discrete frequencies within K-Band. When an intruder enters the detection zone, the Doppler response at each frequency is digitally recorded. The difference between the two frequencies is controlled so that the phase angle between the two Doppler responses can be used to determine the unambiguous range to the target. As a byproduct of the process targets approaching the transceiver can be distinguished from those moving away from the transceiver. Range information is used to enhance the signal processing. The amount of signal integration is increased with range in proportion to the width of the detection pattern so as to optimize the signal to noise ratio (SNR).     

Novel radar target detection

Herein, a novel method of radar target detection based on 2-dimensional (2-D) fractal dimension is proposed. The proposed approach exploits both range information and azimuth information to estimate fractal dimension. Moreover, the approach can increase the number of data points. The above two merits result in the fractal dimension estimated by this method are more accurate and robust than the previous method. The detection performance is also better than the previous, which only makes use of 1-dimensional (1-D) information to estimate fractal dimension. Theoretical analysis and experimental results show that the proposed method performs well in a strong clutter background. The proposed method is also validated by real-life radar data, and the better result has been achieved.     

Bayesian gamma mixture model approach to radar target recognition

This paper develops a Bayesian gamma mixture model approach to automatic target recognition (ATR). The specific problem considered is the classification of radar range profiles (RRPs) of military ships. However, the approach developed is relevant to the generic discrimination problem. We model the radar returns (data measurements) from each target as a gamma mixture distribution. Several different motivations for the use of mixture models are put forward, with gamma components being chosen through a physical consideration of radar returns. Bayesian formalism is adopted and we obtain posterior distributions for the parameters of our mixture models. The distributions obtained are too complicated for direct analytical use in a classifier, so Markov chain Monte Carlo (MCMC) techniques are used to provide samples from the distributions. The classification results on the ship data compare favorably with those obtained from two previously published techniques, namely a self-organizing map and a maximum likelihood gamma mixture model classifier.     

Application of frequency correlation function to radar target detection

Analysis of high-resolution 35 GHz synthetic aperture radar (SAR) imagery of terrain reveals that when point targets, such as vehicles, are viewed at angles close to grazing incidence, they are often difficult to distinguish from tree trunks because the radar cross section (RCS) intensities of the vehicles are comparable to the upper end of the RCS exhibited by tree trunks. To resolve the point target/tree trunk ambiguity problem, a detailed study was conducted to evaluate the use of new detection features based on the complex frequency correlation function (FCF). This paper presents an analytical examination of FCF and its physical meaning, the results of a numerical simulation study conducted to evaluate the performance of a detection algorithm that uses FCF, and the corroboration of theory with experimental observations conducted at 35 and 95 GHz. The FCF-based detection algorithm was found to correctly identify tree trunks as such in over 90% of the cases, while exhibiting a false alarm rate of only 3%.     

A new approach to robust fault detection and identification

A methodology for instrument fault detection and identification (FDI) in linear dynamical systems subject to plant parameter variations or uncertainties is presented. At the heart of this approach is a robust estimator for which the necessary and sufficient conditions to its existence are outlined. The robust estimator can simultaneously estimate the unmeasurable state variables of the system for the purpose of control and provide necessary information for FDI purposes. A novel feature of this approach is that it can actually identify the shape and magnitude of the failures. The scheme allows for fast and accurate FDI, and can account for structural uncertainties and variations in the parameters of the dynamical model of the system. The overall fault tolerant control system strategy proposed is verified through simulation studies performed on the control of a vertical takeoff and landing (VTOL) aircraft in the vertical plane     

基于双隐层过程神经网络的飞机发动机故障检测

利用双隐层过程神经网络模型可以直接处理时变信号的特点，提出了一种用双隐层过程神经网络模型对飞机发动机进行故障检测的方法。由过程神经元隐层完成对输入信息过程模式特征的提取和对时间的聚合运算，非时变一般神经元隐层用于提高网络对系统输入输出之间复杂关系的映射能力。分别利用递归神经网络和双隐层过程神经网络对发动机排气温度裕度进行仿真预测。结果表明，双隐层过程神经网络收敛速度快、精度高，优于递归神经网络的预测结果。为飞机发动机状态监测问题提供了一种有效的方法。     

优选小波神经网络在周跳探测与修复中的应用

北斗导航定位过程中,传统的周跳探测与修复方法缺乏检验环节,无法保证修复结果的可靠性。为此,提出了一种级联式小波变换结合NARX神经网络多步循环预测修复的方法处理周跳问题。该方法通过构造载波相位双差模型检验量,探测周跳发生历元,采用NARX神经网络预测方法修复周跳,利用优选小波基函数进行周跳修复效果检验。实验证明,相较于经验模态分解和变分模态分解等模态分解法,优选小波神经网络周跳探测与修复方法可用于小周跳探测并判断出周跳正负性;构造的NARX神经网络周跳修复模型,解决了普通神经网络模型和传统多项式拟合法容易造成的二次奇异值问题。相较于长短期记忆(LSTM)和门控循环单元(GRU)深度学习神经网络模型,周跳预测精度分别提高了45.2%和55.9%。