Introduction
Food sorting is completed either using manual labor or imaging cameras which provide data to a processor and robotic arm. The imaging cameras allow for the potential to sort a higher amount of material compared to manual labor. At the moment, however, the computational power of traditional processors does not allow for an increase in material compared to manual labor. The goal of the project is to utilize the configurability and computational power of the Field Programmable Gate Array System (FPGA).
The FPGA allows for a customized data flow through the fabric of the transistors. This customized data flow allows for large amounts of data to be processed in real-time. The integrated system will allow for the processing of real-time hyperspectral image data to be used in, but not limited to, the application of food sorting.
Project Background
The hyperspectral imaging system developed by Resonon is widely successful in applications of machine vision sorting due to the ability to see small color differences in an object that conventional imaging systems miss. Currently, using their imaging system, Resonon feeds the data to an actuator system to separate good product from bad. The primary limitation at this point is the speed of processing for the images. To remain at the top of their field, Resonon would benefit from being able to process higher resolution images at faster speeds. The necessary speeds are unattainable with traditional PC processing methods.
Currently in situations utilizing the machine vision technology, there is often still a human working along the line to double-check the machine’s work. Sadly, this is necessary to ensure the sorting is completely effective as the actuator still misses some product. The greatest limitation at this time is in the speed of processing the high-resolution images. Resonon’s imaging system has over 200 data values (corresponding to the spectral channels) per pixel compared to the standard 3 obtained from conventional imaging systems. However, customers would value this technology even more if it could be utilized effectively in real-time. This means that it has the ability to obtain even higher-resolution images and process them at a faster speed to facilitate sorting in real-time. The limitation in this now falls to the computing platform used.