A HIGH-SPEED SORTING SYSTEM FOR DRIED PRUNES

A multi-camera, multiprocessor system was developed to sort prunes for surface defects in real-time. Three line-scan cameras were used to vi ew the prune while in air. Each camera was connected to a subsystem co mputer to process the images simultaneously. Images were stored with a dual memory frame grabber circuit. An algorithm developed previously was modified to analyze and classify the images in less than 15 ms. De fective prunes were removed by a pneumatic system. With a combined sam ple of natural condition prunes containing 28% defective fruit, sortin g errors were 5.6% for good prunes and 10.8% for defective prunes. Err ors for mold, scab, and crack were 3.7%, 9.1%, and 16.5%, respectively . Uneven fruit spacing on the main conveyor and variations in trajecto ry were the main causes of sorting error.