Linux/IWR1443BOOST: Mirrored Ghosts

Hi Rupak,

IN general, ghost detection are caused by multiple reflections from surrounding objects. You can apply certain techniques to try to reduce the probability of ghost detection.

Multiple reflections generally have less intensity as compared to the direct reflections so you can increase the CFAR detection threshold. This may however also reduce the detection performance for genuine weaker objects as well so you'll need to experiment to find an optimum value.

Higher level algorithms such as clustering and tracking can be used with application level knowledge to filter out ghosts. For example, TI's People Counting and Tracking reference design using mmWave Radar Sensor allows you to define the region of interest in the GUI and any points outside that region are ignored/filtered out. It also applies technique in the tracking algorithms to allocate tracks (start tracking and keep tracking) when a group of points is detected continuously across a certain count of frames. While ghosts can be initially detected, this allows the algorithm to quickly reject ghost objects as generally, multipath reflections are transient in nature.

Regards

-Nitin