Other Parts Discussed in Thread: MMWCAS-RF-EVM, IWR6843ISK, AWR2243
Hello,
I have been experimenting with real time object detection using the MMWCAS-DSP-EVM board paired with an MMWCAS-RF-EVM (rev D) board running the Cascade Object Detection sample usecase found in the RADAR SDK (v3.08.0).
I have run the calibration using a corner reflector and generated the calib_coef.bin with the MATLAB script (vision_sdk/apps/src/rtos/radar/src/usecases/cascade_radar_object_detect/create_calib_coefs.m) and am using the default 'radar_test_vector' folder found at (vision_sdk/apps/src/rtos/radar/src/usecases/cascade_radar_object_detect/radar_test_vector).
I've attempted to collect sample point cloud data of a person walking in front of the sensor, however the point cloud does not appear to represent that and instead seems to mainly captured a large number of points a 0 or < 50cm away from the sensor, which can be roughly seen in the below plot of a single captured frame.
Compared to the point cloud frame captured from an IWR6843ISK running the "68xx ISK - People Counting" Lab
Where we can see a point cloud corresponding to a person. I was expecting the sample use case on the MMWCAS-DSP-EVM to output a similar point cloud and would appreciate any guidance on whether this use case would actually be expected to track a person, and if so if there is any advice you can offer on configuring the Cascade radar system to do so.
Thank you,
Robert