Tesis sobre el tema "Micro-Doppler radar"

This thesis reports on research into the field of Micro-Doppler Signature (μ-DS) based radar Automatic Target Recognition (ATR) with additional contributions to general radar ATR methodology. The μ-DS based part of the research contributes to three distinct areas: time domain classification; frequency domain classification; and multiperspective μ-DS classification that includes the development of a theory for the multistatic μ-DS. The contribution to general radar ATR is the proposal of a methodology to allow better evaluation of potential approaches and to allow comparison between different studies. The proposed methodology is based around a “black box” model of a radar ATR system that, critically, includes a threshold to detect inputs that are previously unknown to the system. From this model a set of five evaluation metrics are defined. The metrics increase the understanding of the classifier’s performance from the common probability of correct classification, that reports how often the classifier correctly identifies an input, to understanding how reliable it is, how capable it is of generalizing from the reference data, and how effective its unknown input detection is. Additionally, the significance of performance prediction is discussed and a preliminary method to estimate how well a classifier should perform is developed. The proposed methodology is then used to evaluate the μ-DS based radar ATR approaches considered. The time domain classification investigation is based around using Dynamic Time Warping (DTW) to identify radar targets based on their μ-DS. DTW is a speech processing technique that classifies data series by comparing them with a pre-classified reference dataset. This is comparable to the common k-Nearest Neighbour (k-NN) algorithm, so k-NN is used as a benchmark against which to evaluate DTW’s performance. The DTW approach is observed to work well. It achieved high probability of correct classification and reliability as well as being able to detect inputs of unknown class. However, the classifier’s ability to generalize from the reference data is less impressive and it performed only slightly better than a random selection from the possible output classes. Difficulties in classifying the μ-DS in the time domain are identified from the k-NN results prompting a change to the frequency domain. Processing the μ-DS in the frequency domain permitted the development of an advanced feature extraction routine to maximize the separation of the target classes and therefore reduce the effort required to classify them. The frequency domain also permitted the use of the performance prediction method developed as part of the radar ATR methodology and the introduction of a na¨ıve Bayesian approach to classification. The results for the DTW and k-NN classifiers in the frequency domain were comparable to the time domain, an unexpected result since it was anticipated that the μ-DS would be easier to classify in the frequency domain. However, the naıve Bayesian classifier produced excellent results that matched with the predicted performance suggesting it could not be bettered. With a successful classifier, that would be suitable for real-world use, developed attention turned to the possibilities offered by the multistatic μ-DS. Multiperspective radar ATR uses data collected from different target aspects simultaneously to improve classification rates. It has been demonstrated successful for some of the alternatives to μ-DS based ATR and it was therefore speculated that it might improve the performance of μ-DS ATR solutions. The multiple perspectives required for the classifier were gathered using a multistatic radar developed at University College London (UCL). The production of a dataset, and its subsequent analysis, resulted in the first reported findings in the novel field of the multistatic μ-DS theory. Unfortunately, the nature of the radar used resulted in limited micro-Doppler being observed in the collected data and this reduced its value for classification testing. An attempt to use DTW to perform multiperspective μ-DS ATR was made but the results were inconclusive. However, consideration of the improvements offered by multiperspective processing in alternative forms of ATR mean it is still expected that μ-DS based ATR would benefit from this processing.

As objects move, their changing shape produces a signature that can be measured by a radar system. That signature is called the micro-Doppler signature. The micro-Doppler signature of an object is a distinguishing characteristic for certain classes of objects. In this thesis features are extracted from the micro-Doppler signature and are used to classify objects. The scope of the objects is limited to humans walking and traveling vehicles. The micro-Doppler features are able to distinguish the two classes of objects. With a sufficient amount of training data, the micro-Doppler features may be used with learning algorithms to predict unknown objects detected by the radar with high accuracy.

Thesis (MScEng)--Stellenbosch University, 2014.
ENGLISH ABSTRACT: This work entails the development and performance analysis of a human gait detection system based on radar micro-Doppler signals. The system consists of a tracking functionality and a target classifier. Target micro-Doppler signatures are extracted with Short-Time Fourier Transform (STFT) based spectrogram providing a high-resolution signatures with the radar that is used. A feature extraction mechanism is developed to extract six features from the signature and an artificial neural network (A-NN) based classifier is designed to carry out the classification process. The system is tested on real X-band radar data of human subjects performing six activities. Those activities are walking and speed walking, walking with hands in pockets, marching, running, walking with a weapon, and walking with arms swaying. The multiclass classifier was designed to discriminate between those activities. High classification accuracy of 96% is demonstrated.
AFRIKAANSE OPSOMMING: Hierdie werk behels die ontwikkeling, en analise van werksverrigting, van ’n menslike stapdetekor gebaseer op radar-mikrodoppleranalise. Die stelsel bestaan uit ’n teikenvolger en -klassifiseerder. Die mikrodoppler-kenmerke van ’n teiken word met behulp van die korttyd-Fourier-transform onttrek, en verskaf hoe-resolusie-kenmerke met die radar wat vir die implementering gebruik word. ’n Kenmerkontrekkingstelsel is ontwikkel om ses kenmerke vanuit die spektrogram te onttrek, en ’n kunsmatige neurale netwerk word as klassifiseerder gebruik. Die stelsel is met ’n X-band radar op werklike menslike beweging getoets, terwyl vrywilligers ses aktiwiteite uitgevoer het: loop, loop (hand in die sakke), marsjeer, hardloop, loop met ’n wapen, loop met arms wat swaai. Die multiklas-klassifiseerder is ontwerp om tussen hierdie aktiwiteite te onderskei. ’n Hoe klassifiseringsakkuraatheid van 96% word gedemonstreer.

Les méthodes traditionnelles de formation d'images ISAR supposent que la cible est rigide et ne tiennent pas compte de ses déformations géométriques. Ces mouvements, qui s'ajoutent au déplacement global de la cible, créent une modulation en fréquence sur le signal réfléchi. Ce phénomène, aussi appelé effet micro-Doppler, se traduit d'un point de vue spectral par un étalement des fréquences autour de la fréquence Doppler centrale. Comme les caractéristiques de ces modulations sont directement reliées aux propriétés géométriques et dynamiques de la cible, l'analyse de l'effet micro-Doppler peut apporter des informations complémentaires aux méthodes existantes de reconnaissance de cibles mobiles. Les travaux précédents ont principalement été consacrés à l'analyse temporelle de l'effet micro-Doppler sans tenir compte de la dimension spatiale. En outre, mis à part les cas d'étude théoriques, il existe très peu de modélisations et de données réelles de cibles déformables. A travers les exemples de la roue et du piéton, cette thèse consiste à caractériser finement les effets des déformations géométriques en imagerie radar, en combinant l'analyse en distance et en Doppler. En outre, un accent est mis sur l'influence de la géométrie relative entre le radar et la cible.\\ Ces travaux s'appuient sur un large volet expérimental où sont exploitées les données issues du radar HYCAM, un système d'acquisition large bande développé par l'ONERA. En complément des mesures, le développement d'un outil de simulation permet de faire le lien entre les données réelles et le modèle de l'objet afin d'extraire des grandeurs physiques du phénomène étudié.

I sistemi radar mmWave stanno diventando molto comuni sui veicoli e le loro capacità, in termini di portata e velocità, li rendono adatti a un'altra classica applicazione radar classica, quella relativa all'effetto micro-Doppler. Dall'elaborazione dei segnali radar mmWave, l'effetto micro-Doppler può essere sfruttato, rendendo così possibile estrarre informazioni interessanti sui bersagli. Con l'enorme larghezza di banda e il breve tempo di trasmissione del segnale, l effetto micro-Doppler può essere utilizzato per diversi scopi come la vibrazione del bersaglio o la classificazione dei bersagli. Grazie anche al progresso delle tecniche di Machine Learning, la loro combinazione con elaborazione del segnale radar è un campo interessante da esplorare e può essere usato per fornire soluzioni a diversi problemi radar. L'effetto Micro-Doppler ha una lunga storia nei sistemi radar, un sacco di letteratura può essere trovata su questo argomento, ma la maggior parte di loro considera dispositivi non commerciali quindi è abbastanza lontano da un caso pratico. In questa dissertazione, diverse tecniche per elaborare i segnali micro-Doppler provenienti da radar automobilistici sarà presentato, con lo scopo di classificarli ed estrarre informazioni sulle vibrazioni dal bersaglio. Il contributo principale di questo lavoro è la proposta di nuove tecniche che possono essere applicato su un sensore commerciale e li rende adatti per il micro- Doppler.
mmWave Radar systems are becoming very common on vehicles and their capabilities, in terms of range and velocity, make them suitable for another classical radar application, the one related to the micro-Doppler effect. From the processing of mmWave radar signals, the micro-Doppler effect can be exploited, making so possible to extract interesting information on the observed targets. With the huge bandwidth and the short signal transmission time, the micro-Doppler effect can be used for different purposes such as target vibration measurements or targets classification. Thanks also to the advance of Machine Learning techniques, their combination with radar signal processing is an interesting field to explore and can be used to provide solutions to different radar problems. The Micro-Doppler effect has a long story in Radar systems, a lot of literature can be found on this topic but most of them consider non-commercial devices so is quite away from a practical case. In this dissertation, different techniques to process the micro-Doppler signals coming from automotive radars will be presented, with the purpose of classifying them and extracting vibration information from the target. The main contribution of this work is the proposal of novel techniques that can be applied on a commercial sensor and makes them suitable for the micro- Doppler application.

This Thesis presents radar concepts and signal processing techniques including the fractional Fourier transform (FrFT), Chebyshev polynomial approximation and Singular Spectrum Analysis (SSA) for advanced high resolution radar imaging and micro-Doppler signature extraction. Two novel SAR focussing algorithms in the time-frequency domain are developed using the FrFT. These are called the Fractional Range Doppler Algorithm (FrRDA) and the enhanced Fractional Chirp Scaling Algorithm (eFrCSA. The new methods are tested on simulated and real data sets and are shown to provide higher performances in terms of image quality and resolution than existing frequency domain based methods. The state of the art signal spectrum models of a bistatic point target spectrum for bistatic SAR imaging has been improved by deriving Chebyshev polynomial approximations in place of the conventional Taylor based approximations. This new model increases the accuracy and the efficiency of frequency domain focussing algorithms. Models for micro-Doppler signatures in bistatic SAR are developed and the effect of the different acquisition geometries are considered, including the effect on the final image. A new concept for a Passive Bistatic Radar is introduced for micro-Doppler analysis of helicopters rotor blades. The proposed system exploits the forward scattering enhancement to increase the radar cross section of the helicopters rotor blade allowing an acceptable operative range. The analysis shows how the proposed system could be considered as a good candidate for cheap coast and border control. A detailed analysis on the effect of micro-Doppler from wind turbines and their impact on SAR images is presented. The signal model for such a distributed target is presented and simulation results show how the presence of such a target can significantly decrease and corrupt the image quality. Singular Spectrum Analysis (SSA) is developed for micro-Doppler signature extra ction from SAR clutter and from the direct signal interference and clutter of a passive bistatic radar. The SSA is shown to be robust and capable of performing as a useful tool with the capability of mitigating the effects of clutter on micro-Doppler signatures.

La « vision » à travers les milieux opaques (murs, cloisons, décombres, ou plus généralement tout milieu qui occulte la vision humaine) est l’un des problèmes clefs du contrôle et de la sécurité. Il apparaît à l’heure actuelle un réel besoin de disposer de dispositifs d’observation à travers ces milieux pour des applications tant militaires (lors des assauts, des prises d’otages, etc.) que civiles (recherche de personnes enfouies dans des décombres, dans un incendie, etc). Les avancées sur cette problématique ont conduit à mettre en place des systèmes radars à très courte portée, opérationnels pour la détection et le tracking de personnes dans des environnements simples. Cependant ils nécessitent que les cibles soient en déplacement afin de les différencier des objets statiques. Cette limitation constitue un défaut majeur pour un certain nombre de scénarii réels où des personnes, par stratégie ou par contrainte, restent immobiles. Ces travaux de thèse visent à explorer les mécanismes de détection de personnes statiques par le biais de leurs micro-mouvements, e.g. des mouvements induits par le thorax lors de la respiration. Nous avons étudié - d’un point de vue théorique - les principes physiques sous-jacents à la détection de ces micro-mouvements par radar UWB impulsionnel à partir du mécanisme Doppler impulsionnel. Ce dernier s’appuie sur des mesures consécutives des phases des impulsions réfléchies. La compréhension de ce phénomène a permis de définir une architecture radar impulsionnelle et de la positionner, en termes de contributions, au regard des différents radars UWB proposés dans la littérature : le FMCW et le radar de bruit. Deux dispositifs radars ont servi de support à ce travail. Le premier, de type démonstrateur académique, repose sur l’utilisation d’un oscilloscope rapide pour numériser les impulsions UWB de 3 à 6 GHz de bande. Il a permis de mettre en place une chaîne de traitement complète de vision à travers les murs. Le second dispositif est un prototype radar développé autour d’une plateforme de numérisation ultra-rapide (100 Gsps par échantillonnage équivalent) de fréquence de rafraîchissement très élevée (100 Hz). Il est construit autour d’un FPGA, d’un ADC rapide (1,25 GHz) et d’un T&H très large bande (18 GHz). Il permet ainsi la détection des micro-mouvements par traitement Doppler impulsionnel
"Vision" through opaque environments (walls, partitions, rubble, or any environment that obscures human vision) is one of the key issues of control and security. Advances on this issue have led to operational shortrange radar systems for people detection and tracking in simple environments. However, most of them require the targets to move in order to differentiate them from static objects. This requirement constitues a major shortcoming for a certain number of real scenarios where people, by strategies or by constraints, remain motionless. Hence, this thesis aims to explore the mechanisms of detection of static people through their micro-movements, e.g. movements induced by the thorax during breathing. We have studied - from a theoretical point of view - the physical principles underlying the detection of these micro-movements by pulsed UWB radar with the pulsed Doppler phenomenon, which relies on consecutive measurements of the reflected pulses phases. The understanding of this phenomenon made it possible to define a radar architecture and to position it, in terms of contributions, with regard to the different UWB radars proposed in the literature : the FMCW and the noise radar. Two radar devices served as support for this work. An academic demonstrator based on the use of a fast oscilloscope to digitize the pulses. It allowed to set up a complete processing chain for the application of vision through the walls. The second device is a radar prototype developed around a high-speed scanning platform (100 Gsps perequivalent sampling) with a very high refresh rate (100 Hz). This prototype is built around an FPGA, a fast ADC (1.25 GHz) and a very wide band T&H (18 GHz). This thereby enables to detect micro-movements by pulsed Doppler processing

L’apprentissage automatique, et en particulier l’apprentissage profond, unit un arsenal d’outillages puissants pour modeler et étudier les distributions statistiques sous-jacentes aux données, permettant ainsi l’extraction d’informations sémantiquement valides et interprétables depuis des séquences tabulaires de nombres par ailleurs indigestes à l’œil humain. Bien que l’apprentissage fournisse une solution générique à la plupart des problèmes, certains types de données présentent une riche structure issue de phénomènes physiques: les images ont la localité spatiale, les sons la séquentialité temporelle, le radar la structure temps-fréquence. Il est à la fois intuitif et démontrable qu’il serait bénéfique d’exploiter avec astucieuse ces formations fondatrices au sein même des modèles d’apprentissage. A l’instar des architectures convolutives pour les images, les propriétés du signal peuvent être encodées et utilisées dans un réseau de neurones adapté, avec pour but l’apprentissage de modèles plus efficaces, plus performants. Spécifiquement, nous œuvrerons à intégrer dans la conception nos modèles profonds pour la classification de séries temporelles des sur leurs structures sous-jacentes, à savoir le temps, la fréquence, et leur nature proprement complexe. En allant plus loin dans une veine similaire, l’on peut s’atteler à la tâche d’étudier non pas le signal en tant que tel, mais bel et bien la distribution statistique dont il est issu. Dans ce scénario, les familles Gaussiennes constituent un candidat de choix. Formellement, la covariance des vecteurs de données caractérisent entièrement une telle distribution, pour peu qu’on la considère, à peu de frais, centrée; le développement d’algorithmes d’apprentissage, notamment profonds, sur des matrices de covariance, sera ainsi un thème central de cette thèse. L’espace des distributions diverge de manière fondamentale des espaces Euclidiens plats; il s’agit en fait de variétés Riemanniennes courbes, desquelles il conviendra de respecter la géométrie mathématique intrinsèque. Spécifiquement, nous contribuons à des architectures existantes par la création de nouvelles couches inspirées de la géométrie de l’information, notamment une couche de projection sensible aux données, et une couche inspirée de l’algorithme classique de la Batch Normalization. La validation empirique de nos nouveaux modèles se fera dans trois domaines différents: la reconnaissance d’émotions par vidéo, d’action par squelettes, avec une attention toute particulière à la classification de drones par signal radar micro-Doppler. Enfin, nous proposerons une librairie PyTorch aidant à la reproduction des résultats et la facilité de ré-implémentationdes algorithmes proposés
Machine Learning, and in particular Deep Learning, is a powerful tool to model and study the intrinsic statistical foundations of data, allowing the extraction of meaningful, human-interpretable information from otherwise unpalatable arrays of floating points. While it provides a generic solution to many problems, some particular data types exhibit strong underlying physical structure: images have spatial locality, audio has temporal sequentiality, radar has time-frequency structure. Both intuitively and formally, there can be much to gain in leveraging this structure by adapting the subsequent learning models. As convolutional architectures for images, signal properties can be encoded and harnessed within the network. Conceptually, this would allow for a more intrinsic handling of the data, potentially leading to more efficient learning models. Thus, we will aim to use known structures in the signals as model priors. Specifically, we build dedicated deep temporal architectures for time series classification, and explore the use of complex values in neural networks to further refine the analysis of structured data. Going even further, one may wish to directly study the signal’s underlying statistical process. As such, Gaussian families constitute a popular candidate. Formally, the covariance of the data fully characterizes such a distribution; developing Machine Learning algorithms on covariance matrices will thus be a central theme throughout this thesis. Statistical distributions inherently diverge from the Euclidean framework; as such, it is necessary to study them on the appropriate, curved Riemannian manifold, as opposed to a flat, Euclidean space. Specifically, we contribute to existing deep architectures by adding normalizations in the form of data-aware mappings, and a Riemannian Batch Normalization algorithm. We showcase empirical validation through a variety of different tasks, including emotion and action recognition from video and Motion Capture data, with a sharpened focus on micro-Doppler radar data for Non-Cooperative Target Recognition drone recognition. Finally, we develop a library for the Deep Learning framework PyTorch, to spur reproducibility and ease of use

L’évolution des techniques altimétriques de la bande Ku Nadir vers la bande Ka et l’interféro-métrie large fauchée proche Nadir dans le contexte de la mission SWOT (« Surface Water Ocean Topography », CNES/NASA) soulève de nouvelles questions scientifiques quant à la validité des modèles de rétrodiffusion des surfaces d’eau dans cette bande de fréquence et les erreurs sur les estimations de hauteurs d’eau dues aux mouvements de ces surfaces au cours du temps. Un modèle de rétrodiffusion (GO4) adapté à la configuration SWOT est présenté. Il conserve la précision du modèle de référence de l’Optique Physique tout en gardant la simplicité du modèle plus couramment employé de l’Optique Géométrique. En plus du paramètre classique de pente, il introduit un paramètre supplémentaire, dit de « courbure effective » (msc). Le modèle permet l’inversion des paramètres de pente et de courbure de la surface sous certaines conditions déve-loppées dans ce manuscrit. La validité des modèles conjoints de rétrodiffusion en bande Ka et de surface d’eau a été vérifiée sur des mesures radar effectuées en soufflerie dans un environnement contrôlé. Dans une dernière partie, les propriétés temporelles du signal rétrodiffusé ont été étudiées, en particulier le temps de corrélation et le décalage Doppler induit par le mouvement des vagues. Nous étudions l’influence de ces quantités sur les performances de la synthèse SAR non focalisée du système SWOT
The evolution of the altimetric techniques from Ku-band to Ka-band and the wide swath al-timetry in the context of the SWOT mission (« Surface Water Ocean Topography », CNES/NASA) raises new scientific questions about the validity of the backscattering models from water sur-faces in such a frequency band and errors in estimating water heights from time-evolving water surfaces. A backscattering model (GO4) adapted to the SWOT configuration is introduced. It preserves the accuracy of the referencial Physical Optics model while maintaining the simplicity of the clas-sical Optical Geometrics model. In addition to the classical slope parameter, it introduces another parameter called as « effective curvature » (msc). This model allows the inverson of the surface slope and curvature parameters under certain conditions which are developped in this manus-cript. The joint validity of the backscattering models in Ka-band and from water surfaces is che-cked from controlled wind-wave tank radar measurements . In a last part, the temporal properties of the backscattered signal is studied, in particular the correlation time and the Doppler shift induced by waves motion. Influence of the latters on the non focused SAR synthesis is studied in the context of the SWOT system

L’évolution des techniques altimétriques de la bande Ku Nadir vers la bande Ka et l’interféro-métrie large fauchée proche Nadir dans le contexte de la mission SWOT (« Surface Water Ocean Topography », CNES/NASA) soulève de nouvelles questions scientifiques quant à la validité des modèles de rétrodiffusion des surfaces d’eau dans cette bande de fréquence et les erreurs sur les estimations de hauteurs d’eau dues aux mouvements de ces surfaces au cours du temps. Un modèle de rétrodiffusion (GO4) adapté à la configuration SWOT est présenté. Il conserve la précision du modèle de référence de l’Optique Physique tout en gardant la simplicité du modèle plus couramment employé de l’Optique Géométrique. En plus du paramètre classique de pente, il introduit un paramètre supplémentaire, dit de « courbure effective » (msc). Le modèle permet l’inversion des paramètres de pente et de courbure de la surface sous certaines conditions déve-loppées dans ce manuscrit. La validité des modèles conjoints de rétrodiffusion en bande Ka et de surface d’eau a été vérifiée sur des mesures radar effectuées en soufflerie dans un environnement contrôlé. Dans une dernière partie, les propriétés temporelles du signal rétrodiffusé ont été étudiées, en particulier le temps de corrélation et le décalage Doppler induit par le mouvement des vagues. Nous étudions l’influence de ces quantités sur les performances de la synthèse SAR non focalisée du système SWOT
The evolution of the altimetric techniques from Ku-band to Ka-band and the wide swath al-timetry in the context of the SWOT mission (« Surface Water Ocean Topography », CNES/NASA) raises new scientific questions about the validity of the backscattering models from water sur-faces in such a frequency band and errors in estimating water heights from time-evolving water surfaces. A backscattering model (GO4) adapted to the SWOT configuration is introduced. It preserves the accuracy of the referencial Physical Optics model while maintaining the simplicity of the clas-sical Optical Geometrics model. In addition to the classical slope parameter, it introduces another parameter called as « effective curvature » (msc). This model allows the inverson of the surface slope and curvature parameters under certain conditions which are developped in this manus-cript. The joint validity of the backscattering models in Ka-band and from water surfaces is che-cked from controlled wind-wave tank radar measurements . In a last part, the temporal properties of the backscattered signal is studied, in particular the correlation time and the Doppler shift induced by waves motion. Influence of the latters on the non focused SAR synthesis is studied in the context of the SWOT system

abstract: The micromotions (e.g. vibration, rotation, etc.,) of a target induce time-varying frequency modulations on the reflected signal, called the micro-Doppler modulations. Micro-Doppler modulations are target specific and may contain information needed to detect and characterize the target. Thus, unlike conventional Doppler radars, Fourier transform cannot be used for the analysis of these time dependent frequency modulations. While Doppler radars can detect the presence of a target and deduce if it is approaching or receding from the radar location, they cannot identify the target. Meaning, for a Doppler radar, a small commercial aircraft and a fighter plane when gliding at the same velocity exhibit similar radar signature. However, using a micro-Doppler radar, the time dependent frequency variations caused by the vibrational and rotational micromotions of the two aircrafts can be captured and analyzed to discern between them. Similarly, micro-Doppler signature can be used to distinguish a multicopter from a bird, a quadcopter from a hexacopter or a octacopter, a bus from a car or a truck and even one person from another. In all these scenarios, joint time-frequency transforms must be employed for the analysis of micro-Doppler variations, in order to extract the targets’ features. Due to ample bandwidth, THz radiation provides richer radar signals than the microwave systems. Thus, a Terahertz (THz) micro-Doppler radar is developed in this work for the detection and characterization of the micro-Doppler signatures of quadcopters. The radar is implemented as a continuous-wave (CW) radar in monostatic configuration and operates at a low-THz frequency of 270 GHz. A linear time-frequency transform, the short-time Fourier transform (STFT) is used for the analysis the micro-Doppler signature. The designed radar has been built and measurements are carried out using a quadcopter to detect the micro-Doppler modulations caused by the rotation of its propellers. The spectrograms are obtained for a quadcopter hovering in front of the radar and analysis methods are developed for characterizing the frequency variations caused by the rotational and vibrational micromotions of the quadcopter. The proposed method can be effective for distinguishing the quadcopters from other flying targets like birds which lack the rotational micromotions.
Dissertation/Thesis
Masters Thesis Electrical Engineering 2018

Radar (RAdio Detection And Ranging) uses radio waves to detect the presence of a target and measure its position and other properties. This sensor has found many civilian and military applications due to advantages such as possible large surveillance areas and operation day and night and in all weather. The contributions of this thesis are within applied signal processing for radar in two somewhat separate research areas: 1) radar with array antennas and 2) radar with micro-Doppler measurements. Radar with array antennas: An array antenna consists of several small antennas in the same space as a single large antenna. Compared to a traditional single-antenna radar, an array antenna radar gives higher flexibility, higher capacity, several radar functions simultaneously and increased reliability, and makes new types of signal processing possible which give new functions and higher performance. The contributions on array antenna radar in this thesis are in three different problem areas. The first is High Resolution DOA (Direction Of Arrival) Estimation (HRDE) as applied to radar and using real measurement data. HRDE is useful in several applications, including radar applications, to give new functions and improve the performance. The second problem area is suppression of interference (clutter, direct path jamming and scattered jamming) which often is necessary in order to detect and localize the target. The thesis presents various results on interference signal properties, antenna geometry and subarray design, and on interference suppression methods. The third problem area is measurement techniques for which the thesis suggests two measurement designs, one for radar-like measurements and one for scattered signal measurements. Radar with micro-Doppler measurements: There is an increasing interest and need for safety, security and military surveillance at short distances. Tasks include detecting targets, such as humans, animals, cars, boats, small aircraft and consumer drones; classifying the target type and target activity; distinguishing between target individuals; and also predicting target intention. An approach is to employ micro-Doppler radar to perform these tasks. Micro-Doppler is created by the movement of internal parts of the target, like arms and legs of humans and animals, wheels of cars and rotors of drones. Using micro-Doppler, this thesis presents results on feature extraction for classification; on classification of targets types (humans, animals and man-made objects) and human gaits; and on information in micro-Doppler signatures for re-identification of the same human individual. It also demonstrates the ability to use different kinds of radars for micro-Doppler measurements. The main conclusion about micro-Doppler radar is that it should be possible to use for safety, security and military surveillance applications.

Micro-Doppler is defined as scattering produced by non-rigid-body motion. This dissertation involves the design of a multiple frequency continuous wave (MFCW) radar for micro-Doppler research and detection and classification algorithm design. First, sensor hardware is developed and tested. Various design tradeoffs are considered, with the application of micro-Doppler based detection and classification in mind. A diverse database of MFCW radar micro-Doppler signatures was collected for this dissertation. The micro-Doppler signature database includes experimental data from human, vehicle, and animal targets. Signatures are acquired from targets with varying ranges, velocities, approach angles, and postures. The database is analyzed for micro-Doppler content with a focus on its application to target classification. Joint time-frequency detection algorithms are developed to improve detection performance by exploiting noise-spreading and the micro-Doppler phenomenon. Following detection algorithm development, this dissertation covers the design of micro- Doppler feature extraction, feature selection, and classification algorithms. Feature selection is performed automatically via a Fisher score initialized sequential backward selection algorithm. Classification is performed using two distinct approaches: a generative statistical classification algorithm based on Gaussian mixture models (GMMs) and a discriminative statistical classification algorithm based on support vector machines (SVMs). Classifier performance is analyzed in detail on a micro-Doppler signature database acquired over a three-year period. Both the SVM and GMM classifiers perform well on the radar target classification task (high accuracy, low nuisance alarm probability, high F-measure, etc.). The performance of both classifiers is remarkably similar, and neither algorithm dominates the other in any performance metric when using the chosen feature set. (However, the difference between SVM and GMM classification accuracy becomes statistically significant when many redundant features are present in the feature set.) The accuracy of both classifiers is shown to vary as a function of approach angle, which physically corresponds to the angular dependence of micro-Doppler. The results suggest that overall classifier performance is more sensitive to feature selection than classifier selection (with GMM being more sensitive to redundant features than SVM). Both classifiers are robust enough to handle human targets attempting to evade detection by either army crawling or hands-and-knees crawling.
text

Radar sensors hold excellent capabilities to estimate distance and motion accu- rately, penetrate nonmetallic objects, and remain unaffected by weather conditions. These capabilities make these devices extremely flexible in their applications. Elec- tromagnetic waves centered at frequencies around 24 GHz offer high precision target measurements, compact antenna and circuitry design, and lower atmospheric absorp- tion than higher frequency-based systems. This thesis presents a case study for a 24 GHz frequency modulated continuous wave radar module. We start by addressing the theoretical background necessary for this work and describing the architecture of the module used. We present three classes’ classification accuracy, namely pedes- trians, cyclists, and cars. A set of features for the classification is designed based on theoretical models, and their effectiveness is validated through experiments. The features are extracted from the available geometrical and motion-related information and used to train different classification models to compare the results. Finally, a trade-off between feature number and accuracy is presented.

碩士
國立中央大學
遙測科技碩士學位學程
102
Radar Doppler has various applications in both military and civil aspect. The target moving parameters such as moving direction, moving velocity and so on can be well estimated through the analysis of radar Doppler effects. A new research called micro-Doppler which is proposed by the famous scholar Victor C. Chen was developed on the basis of micro motion model of objects. Since the human beings and the other creatures usually walk or run with a constant velocity, stride and gait frequency, there are some micro-Doppler signatures on their spectrums. Human gait detection and characteristic extraction mainly utilize the human’s walking dynamic features to modulate the radar echo phase, use time-frequency analysis method to obtain the spectrum, and extract the micro-Doppler signature of human gait from spectrum. Through wall surveillance (TWS) is a kind of detection technology which can penetrate walls and other obstacles. With the ability of through-obstacle detection, TWS has been playing an important role in military reconnaissance, counter-terrorism, hostage rescue, earthquake casualty search and so on. In this study, the micro-Doppler effect of the vibration and rotation is firstly discussed which is depend on the specific micro motion, and followed by the tool of time-frequency analysis. A gait echo model is proposed, which represents the rotation angles, displacements and relative phase relationship of joints, scatterers and the instantaneous Radar Cross Section (RCS). Finally, micro Doppler is combined with TWS for the feasibility of human gait direction determination.