Revision history [back]

Basically, if your environment is constant, you could get a long way with the combination of:

• Background segmentation to identify the moving buns. Since the conveyor belt stays same color, this can be neglected as a moving part. From camera vieuwpoint this is a fixed color region.
• Create blobs for each bun
• Track the blobs and assign a counter to the amount of blobs passing a line or so

But again, this is very basically what you should do.

Other possibility could be performing object detection for variable object classes and object recognition for constant object size and look.

OpenCV and C++ work good together on embedded platforms, but that's a choice you will need to make yourself :)