Application of parallel computing in image processing for grading of citrus fruits

This paper describes the development activity on a low cost high performance fruit grader employing distributed network architecture to interface the GigE machine vision camera and field IO devices. The fruits are transported over a dynamic conveyor and images are synchronously captured for frame analysis during triggered events. Geometrical features such as diameter, perimeter, area and circularity are extracted under a laboratory simulated real-time condition in the absence of suitable field conveyor. Task Parallel Library data flow computing concept is adopted under windows .Net Framework for effective individual core resource utilization in a multi core CPU configuration in order to achieve high performance measure. The scalable and upgradable dual core PC based hardware components are utilized for direct interface to cater the individual application demands by suitable software configuration. The system is capable of processing 6 to 10 fruits per second throughput per lane with good measurement accuracy and repeatability.