People counting and tracking with FMCW radar for augmented video surveillance systems

• Autor(a) principal: Santos, Bruno Miguel Silva
• Data de Publicação: 2021
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10773/33916
• topic: Radar; People detection; Sensor fusion; Tracking

The present dissertation describes how a frequency-modulated continuous-wave (FMCW) Radar can be employed to count and locate persons inside a room. Apart from counting, it is also performed the spatial person tracking through processing of the Radar point cloud. A theoretical study is done, detailing how the Radar can compute fundamental parameters like distance, velocity, and angle of arrival of an object present in the field of view (FoV) and the signal processing computations done with the selected Radar evaluation kit. The application developed in the scope of this dissertation contemplates topics such as, noise mitigation, clustering algorithms and association with detected persons and/or objects. It also uses digital signal processing (DSP) algorithms capable of movement prediction in noisy environments with data extracted from the Radar and approaches the projection of Cartesian coordinates into a video surveillance camera located at a different place. This project was intended for security and safety systems, wherein the Radar would be integrated in a video surveillance camera, complementing people detection capabilities achieved via image processing algorithms, which fail in certain atmospheric or lightning conditions while the Radar is less affected. The graphical user interface (GUI) application developed in this dissertation was tested in scenarios where people would walk in regular patterns or are seated in their workplace and constructed in a way that allows it to work in real-time without major losses on the video acquisition performance. Practical results of diverse person detection and tracking use cases are demonstrated with an interesting performance.