On a custom FPGA camera, an algorithm in VHDL/C has been implemented to classify bulk good by color, material and size in the glass recycling area. These data can be used in a further step to sort the different materials.
The centerpiece of the camera forms an Altera Cyclone FPGA, which reads and processes the data from an IBIS4 matrix sensor. Among other things, digital filters are used to improve the quality of the data and to enable the transformations in various color spaces. The developed VHDL modules use a generic data bus, so that their positioning in the signal path can be varied to achieve optimal performance. The Avalon bus enables the modules to be configured and read via the implemented NIOS softcore processor on the Cyclone.
The object recognition algorithm, that is based upon the sensor and preprocessed data, is implemented in C and very flexible configurable. The detected objects can be readout via a serial data bus to drive a sorting unit.