A Thresholding Algorithm for Radar-Based Object Detection

Abstract

In this dissertation, we propose a thresholding algorithm for radar-based object detection. First of all, the researches of detection algorithms are described. Then, two design examples of radar-based object detection are introduced. One is a direction recognition algorithm for detecting the moving direction of people, and the other one is a bi-directional passing people counting system to count the number of passing people in two directions. Finally, a thresholding algorithm for radar-based object detection is described. The direction mentioned in direction recognition algorithm is detected in two ways, in-bound direction and out-bound direction. It is single radar-based direction recognition algorithm which is basically realized by detecting whether the moving direction is towards to the radar or far away from the radar. Therefore, it is normally used in a narrow passageway or a door due to the limitation of the detection range. To extend the detection range, a two radars-based bidirectional passing people counting system is introduced. The system is designed as one embedded module with two radar sensors and one Raspberry Pi 2. The moving direction is recognized based on the time difference that detected in two radars. In order to verify the performance of the bi-directional passing people counting system, we installed the embedded module in a passageway at Gangnam subway station, and compared the measured results with the actual number of passing people. Then, a thresholding algorithm for radar-based object detection is described. Unlike the conventional thresholding algorithm which is based on the noise signal, the proposed algorithm uses both noise and target signals in designing a threshold. So, a threshold can be designed based on the focus point. The focus point represents to the false alarm issue or miss-detection issue. Also, both false alarm and miss-detection rates for the designed threshold can be estimated. In order to verify the performance of the algorithms proposed in this dissertation, we conducted various of experiments using the commercial impulse radio ultra wideband (IR-UWB) radar, and the results are presented in the dissertation.