1 reply to this topic

[jacquesfarges]

Some image processing applications require multiple cameras, and it can be useful to preprocess each image stream to allow a single processor to handle the final processing for all cameras. One way to acheive this would be to develop a hardware device with 2 Ethernet ports, that would be attached to any fast off-the-shelf ethernet-based camera, and which would create a second stream of ethernet data in parallel with stream of images. The stream of images could either be left unaltered, or could be enhanced in some way. I refer to this device as a "filter", drawing a parallel to DirectShow filters. A development environment would provide me with an empty software module that imports a stream of images in groups of rows of pixels, and exports 2 streams, a stream of pixels and a stream of preprocessor result data. A user could define code to fill the software module using ImpulseC to provide custom preprocessing, and then the development environment would used to load an image to the FPGA in the hardware filter, which would be ready to use.

I think this could be an interesting product for someone. Using this for image preprocessing would be only one possibility, since it's really a generic ethernet filter. To use it for image processing, on the FPGA fabric IP would need to be added to convert typical camera ethernet data to a pixel stream and then back again.

[etrexel]

...
device with 2 Ethernet ports, that would be attached to any fast off-the-shelf ethernet-based camera, and which would create a second stream of ethernet data in parallel with stream of images. The stream of images could either be left unaltered,
...
I think this could be an interesting product for someone. Using this for image preprocessing would be only one possibility, since it's really a generic ethernet filter. To use it for image processing, on the FPGA fabric IP would need to be added to convert typical camera ethernet data to a pixel stream and then back again.

Hi,
What specific type of Ethernet cameras are you envisioning would be used? I ask because most off-the-shelf surveillance type cameras with Ethernet connections encode the image into some format like MPEG and sends the data via TCP/IP, both of which are considerable in complexity.

Best Regards,
Ed