In addition to the high-resolution, hardware-assisted tracking of objects on the basis of color, the Cognachrome vision system is capable of low-resolution, 24 bit RGB frame grabs. Standard frame grab resolution is 64x48 pixels.
Although the board's standard software is capable of grabbing frames of video and uploading them to a host for display, frame grabbing is mainly intended for debugging, or for use in systems where the user writes his own software on the board to process the frames. (The communication overhead is simply too great to have good performance in a system which must send every pixel over the serial port.)
If you are using the Pioneer vision system software, you can grab and display a frame at 64x48 resolution from Saphira prompt using the `G' command. See section Saphira Interaction Mode, for details.
To grab and display a frame at 64x48 resolution from the TPUVIS Prompt, do the following:
tpuvis[a]> c Enter advanced color menu colorplot> g Grab a frame into buffer colorplot> u Upload a frame in hex over serial port
The frame data will now be uploaded over the serial port.
colorplot> q Exit advanced color menu tpuvis[a]>
The frame information is sent over stream 30 in the following format:
V 64 48 rgb 64 cols, 48 rows, R, G, and B data
H 0 675a5b 6a5c5b ... 2f2f1b Horizontal line #0, RGB data
H 1 6a5c5b 675a4f ... 2f2f1b Horizontal line #1, RGB data
...
H 47 604222 5a3f22 ... 0f0d05 Horizontal line #47, RGB data
Z End of frame
See section Programming the Vision System.
The hardware limitation of 64 pixels per line can be overcome in two ways: interleave, and removing color channels.
In order to interleave, one divides the Y resolution by two to double the X resolution. Since the Y resolution can be as high as 250 (every single line), this often isn't a major problem. This interleave is performed by grabbing adjacent horizontal lines with a half-pixel offset. The two adjacent horizontal lines are then merged into a single horizontal line. Resolutions as high as 130 x 125 can be obtained using this method.
The hardware Cognachrome Vision System is capable of doing somewhat higher resolution grabs if you are willing to give up color channels. The first step is to move from 24 bits to 16 bits per pixel, which allows you to grab either red and green, or blue and green for each pixel. The second step would be to choose a single color channel (red, green, or blue) and only take 8 bits per pixel.
Note that in 8-bit mode the board can effectively grab a black-and-white image by simply reducing the color saturation adjustment on the NTSC to RGB daughterboard, and then grabbing one of the three channels.
None of the scenarios described below have been implemented in the current software.