Exact Difference between different eVideoModes

[Anzak]

Hello,
I am working on testing a tracking and reconstruction algorithm. I want to know what exactly is the difference between 8 eVideoModes of the camera. Plus 4 of them gives apparently the same result.

Getting grayscale and objects simultaneously will help me a lot, if it is not possible, please help me by telling the precise difference between object, segment, MJPEG and precision modes. I know they are not alike but I need to know the technical difference.

Thanks in anticipation
Anzak

[NaturalPoint - Mike]

Here is a basic rundown of the primary functioning modes of the cameras. If you have more specific requests, I'll have to pass them along to see what information we can give you regarding the specific functionality of the software.


Camera Processing modes
These options select the in-camera video pre-processing mode.

Precision Greyscale: Most precise marker detection, delivering thresholded grayscale marker blobs to the PC for calculation of object location, size, and roundness using our most precise processing methods. Provides greater accuracy than other modes, but consumes more CPU resources and USB bandwidth. It is recommended that cameras operating in this mode have unwanted light sources carefully blocked/masked during calibration in order to prevent excessive USB bus traffic. This is our recommendation for marker tracking in most cases.

Segment: Similar to Precision Grayscale, Segment processing streamlines marker information for more CPU-efficient marker detection. Provides a balance between precision and USB bandwidth use. Can be useful if your PC struggles with Precision Grayscale.

Object: The most CPU-efficient marker detection, as the majority of the image processing takes place on the camera. Provides more efficient USB bandwidth utilization, but delivers less marker information to the PC than other modes. In most cases, only use if your PC is too underpowered for Precision Grayscale or Segment. Doesn't deliver marker blobs--instead, 2D marker locations are calculated on the camera and sent to the PC.

MJPEG-compressed Grayscale: Full resolution grayscale frames, compressed on-camera for scalable scene video capabilities. Enables the use of full-frame images from multiple cameras at once, without increasing bandwidth or CPU footprint as heavily as Raw. 10% the bandwidth as Raw.

Raw Grayscale: Full resolution, uncompressed grayscale.

[LinusA]

Let me just add my experience to what Mike said:

Precision Grayscale:: The cam seems to do a thresholding of the pic (i.e. the threshold property is used). The thresholded 1-bit picture is used as a mask over the cam's internal grayscale view, and you get the masked data to your PC. Basically the shape of blobs, but with real grayscale values in it. Marker positions seem to be calculated as weighted centroids for each blob.

Segment:: In this case you seem to get the 1 bit thresholded picture (you can however set the color of the thresholded "positive" values, so in reality you get a 256-bit 1 channel picture as before). Segmented means the regions are found for you (connected pixels). The marker centroids are obviously less precise, as the grayscale information is lost.

Object:: The cam does internally the same as for Precision Grayscale or Segment modes, but you only get the marker data back to the computer. When you want the actual image, it seems to be rendered by the Camera SDK -- so for a blob, a filled circle will be put at it's location with the radius calculated on the cam. You don't have a real image.


Please note: Precision Grayscale and Segment mode can overflow (not sure about object).
If there are too many thresholded pixels (i.e. if your threshold value is far too low), then the cam just "cuts off" the pixels when the transfer rate exceeds ~1800 KiB/s on my system. I think this value can be adjusted somewhere. Anyway, when this happens, the picture is probably not usable anyway (far too many markers, or too bright / too high exposure).


I'd recommend you either play around with OptiShow, or you write your own test application to try the different modes. I know it helped me a lot

[Seth Steiling]

[quote=LinusA]Object:: The cam does internally the same as for Precision Grayscale or Segment modes, but you only get the marker data back to the computer...[/quote]
One quick note: The V100:R2 and V120:SLIM Object mode uses the Segment approach on-board, while the S250e Object mode uses the Precision Grayscale approach on-board.