AWR1843BOOST: Increasing Point Cloud Density for Human Shape Detection – AWR1843

Tool/software:

Hello,

I am working on classifying detected objects as human or non-human using the AWR1843 EVM. To achieve this, I am extracting point cloud data and attempting to visualize object shapes (essentially, radar cube plotting).

Setup Details

• I am using DSS and MSS applications.

• The DPC at DSS is utilizing the Object Detection DPU.

• I collected a point cloud dump for a scenario where humans were standing in front of the radar for approximately 1 minute at a range of less than 2 meters.

Issue Faced

• Upon plotting the point cloud, I observe that most points with zero elevation.

• The sparse detection makes it difficult to approximate the shape of a human, which I need for training an ML model for human classification.

I am attaching the dump file and the radar configuration used for reference.

Request for Assistance

1. How can I increase the number of detected points to obtain a denser point cloud that better represents the human shape?

2. Are there specific radar configuration parameters (e.g., chirp/profile settings, CFAR thresholds, detection settings) that should be optimized for better point cloud resolution?

3. Any recommended post-processing techniques to refine the captured data?

Any guidance on improving detection density and achieving a more detailed human shape representation would be greatly appreciated!

Thanks in advance!

Attachments

1. Point Cloud Dump

2311.data.txt

2. Radar Config

awr1843_config.cfg