The information below describes how the “CFAR point cloud” output from a radar is derived from the radar’s view of the environment surrounding it.
This mode is just one approach to extracting features from the scene that the radar presents.
The key parameters to configure the operation of this feature extraction process can be configured in a radar by Colossus message type 205 and the navigation mode output from the radar can be requested by Colossus message type 120. When this feature extraction method is appropriate, the network bandwidth utilisation for the radar’s connection is greatly reduced.Alternatively, the source code for navigation mode feature extraction function is provided for both C++ and Pythonenvironments. This approach can be used to process the raw radar data within a customer’s complete control, and dynamic adjustment/adaptation of the key parameters is possible.
| Info |
|---|
|
...
International Research Journal of Engineering and Technology (IRJET) Volume: 05 Issue: 07 describes Constant False Alarm Rate (CFAR) Detection as
… an adaptive algorithm used in Radar systems to detect the target echoes against a background of noise and clutter. The role of the constant false alarm rate circuit is to determine the threshold above which any returning signal or echo can be considered probably to be originated from a target. In most radar systems, the threshold is set to achieve a required false alarm rate (or equivalently, probability of false alarm). Cell-Averaging CFAR (CA - CFAR) is a type of CFAR detection where the threshold is estimated by scanning a block of cells around a cell-under-test (CUT) and calculating the average power level. In CA – CFAR, a target is declared to be present if the power level in the CUT exceeds the average power level found from adjacent block of cells. This paper shows the principle of CA – CFAR detector, threshold factors for CFAR detection, factors affecting CFAR detection and CFAR loss
The subsequent text of that article expands on the technicalities around Cell-Averaging-CFAR.
In very simplistic terms, a single azimuth of radar data is processed. A sliding window approach identifies whether the central datapoint within the window is notably larger (in power terms) than the mean value of data within the window. If the central datapojnt exceeds the local mean by a defined threshold, then the point is considered to be a valid object within the radar’s view.
Note that In order to increase the capability & sensitivity of this approach, (n) datapoints either side of the central point in the window are NOT considered when calculating the local mean.
| Info |
|---|
Radar firmware versions 3.1.0.325 and later support the generation of a Cell-Averaged-CFAR “pointcloud” data output format via a UDP stream |
| Note |
|---|
A radar must be configured via its' web interface to output data in this mode |
| Info |
|---|
Note that while CA-CFAR processing can be carried out within the radar, the source code to execute this processing within a client application outside the radar is also available from our Bitbucket repository. |
...
Cell-Averaging CFAR feature extraction:
Pictorial guide to processing steps
The images below have been captured from a simple CA-CFAR calculation sheet, working with a single azimuth of raw radar output. The calculation sheet can be downloaded here:
| Attachments | ||||
|---|---|---|---|---|
|
...
...
...
...
...
