Person Detection Using Ultra-Wideband Radars

Abstract

In this report, an indoor target detection system is developed for detecting one or multiple targets in a cluttered area based on a network of distributed ultra-wideband radars. The system is capable of acquisition of data in the form of distances between radar and targets, which can be used to localise and track one or multiple targets in real-time. The system can be used in a wide range of applications, ranging from security, such as anti-intruder systems, to commercial applications, such as tracking animals in a zoo.
ewline

A network of four Time Domain PulsON 410 ultra-wideband radars are used in the sensing system, due to their availability, high spatial accuracy and performance in non-line-of-sight conditions. By using different transmission codes, interference between the sensors is kept to a minimum. Multiple antennas are considered, and the Vivaldi antenna is shown to be preferred over the dipole antenna for this application due to its slightly higher directivity, reducing the impact of multi-path effects. The data acquisition for the four radars is initiated almost simultaneously, up to the point that the radars can be considered synchronised. The received signal is filtered using background rejection and an FIR motion filter, to reveal motion of the target. Four detection algorithms are considered, of which the least-of constant false alarm rate (LO-CFAR) has a better performance for indoor multiple person tracking applications over conventional CFAR detection. The LO-CFAR algorithm can be applied to the filtered signal to detect targets in range of the sensors, and determine their distance from the sensor. Range-Doppler processing is proposed as a method to acquire a velocity estimate of the target, which can be used in a tracking system to estimate the position of the target more accurately. However, due to the relatively low slow-time sampling frequency, it is not feasible to utilise Range-Doppler processing in the current system. Only simulations of Range-Doppler processing are shown in the thesis.
ewline

The system is tested in different environments; single person and multiple person situations are considered in an open and a cluttered area. It is shown that the developed system is capable of detecting a single target in both open and cluttered areas. With two targets in cluttered areas, it becomes difficult to distinguish a target from multi-path reflections, reducing the reliability of the system in these situations. In an open area, multiple targets can successfully be distinguished and detected. The sensing system can successfully detect and determine the distance to a person in non-line-of-sight conditions. Additionally, the detection system is tested on a smaller target to determine performance and accuracy of the system in situations with smaller targets. While the small target is possible to detect, the range at which reliable results are obtained is significantly reduced by the size of the target.
ewline

It is concluded that there is room for improvement of the target detection system, especially in situations involving multiple targets in cluttered areas. Through-wall detection is shown to be feasible with the current sensing system. For small targets, a higher pulse integration index is required to achieve a reliable range similar to the results of person detection. Range-Doppler processing requires a higher slow-time sampling frequency in order to be feasible in a real-time tracking system. Further recommendations include testing alternative antennas to improve through-wall detection, designing a multi-static radar system and implementing more complex detection algorithms that have a better multi-person performance.