Advanced Target discrimination using fingerprinting based on High Resolution Range and micro-Doppler profiles

陆基机载成像雷达系统经常需要提供超越目标探测和跟踪的增强态势感知信息。特别是，目标识别是现代雷达经常被要求提供的一项具有挑战性的重要任务。了解目标的特征，包括轮廓、极化响应和微运动，是能够表征和区分合作和非合作目标（如车辆、弹道导弹和船舶）的基础。新型雷达系统现在能够提供丰富的目标信息，例如高分辨率距离轮廓（hrrp）和微多普勒（MD）特征，然而，关于如何利用和最大化这些系统对目标识别任务的好处的研究报告要少得多。该项目将开发模型和算法，以利用和评估从HRRP和MD中提取的我们称之为“目标指纹”的能力，以执行高级目标识别。将利用不同目标的微运动特征来提取独特的特征，从而在多个类别之间进行准确的区分。特别令人感兴趣的是目标真实运动模型的推导以及不同运动和观测参数（例如机动和观测角度）影响接收雷达回波的方式。距离和多普勒域的个体特征将结合起来，并通过新的信号处理方法生成详细的、类似图像的表示和更抽象的特征向量。目的：开发模型，信号处理解决方案和系统（例如波形），以便通过利用高分辨率距离和微多普勒剖面的融合来区分目标。-了解高分辨率范围和微多普勒剖面和特征提取的原理；-发展联合模型，以表示目标的距离-多普勒；-开发目标指纹识别的监督和无监督（深度学习）框架；-开发多目标场景下基于目标指纹的目标跟踪算法；-开发基于指纹和追踪（生活模式）的目标行为预测算法；-利用斯特拉斯克莱德的雷达传感器在受控环境中获取实验数据；-在实验室数据上验证模型和算法；-根据Leonardo提供的数据验证模型和算法；

Ground-based airborne imaging radar systems are frequently required to provide enhanced situational awareness information beyond target detection and tracking. In particular, target recognition is an important challenging task that modern radars are frequently requested to provide. Understanding the characteristics of targets, including profiles, polarimetric responses and micro-motions is fundamental to be able to characterise and discriminate cooperative and non-cooperative targets such as vehicles, ballistic missiles and ships. Novel radar systems are now able to provide enriched target information, such as High Resolution Range Profiles (HRRPs) and micro-Doppler (MD) signatures, however there has been much less research reported on how to exploit and maximise the benefits of these systems for the target recognition task. This project will develop models and algorithms to exploit and assess the capabilities of what we will call " a target's fingerprint" extracted from both HRRP and MD to perform advanced target recognition.Micro-motion characteristics of different targets will be utilised in order to extract unique signatures leading to accurate discrimination between a number of classes. Of particular interest is the derivation of realistic motion models of targets and the way that different motion and observation parameters, e.g. manoeuvring and observation angle, affect the received radar returns. Individual characteristics in range and Doppler domain will be combined and used to generate both detailed, image-like representations and more abstract feature vectors through novel signal processing approaches.Aim: Develop models, signal processing solutions and systems (e.g waveforms) in order to discriminate targets by exploiting a fusion of High Resolution Range and micro-Doppler profiles.Objectives:- Understand the principles of High Resolution Range and Micro-Doppler profiles and signature extraction;- Develop joint models for Range-Doppler representation of targets;- Develop supervised and unsupervised (Deep Learning) frameworks for target fingerprinting;- Develop algorithms for target fingerprinting based target tracking in multi-target scenario;- Develop algorithms for target behaviour prediction based on fingerprinting and tracking (patterns of life); - Acquire experimental data in controlled environment using radar sensors available at Strathclyde; - Validate models and algorithms on lab data; - Validate models and algorithms on data provided by Leonardo;