Sorting of items on a moving conveyor belt. Part I. a technique for detecting and classifying objects

In many situations of industrial interest, a sorting process has to be perf ormed in order to separate different or undesired items among those flowing on a moving conveyor belt. This process consists of looking at the items d istributed on the conveyor, localizing any single item, classifying it on t he basis of features that make it acceptable or not and, if applicable, gri pping it to perform the necessary separation. Fundamental issues here are t herefore se,lsing, i.e., detecting and classifying the items, and gripping, i.e., realizing the required separation in the most efficient way. The two parts of this work deal with these two issues in automatic sorting and pre sent advanced solutions for the sorting of recyclable packaging, an applica tion where the unstructured work environment represents the main difficulty towards the process automation. In Part I. in particular; the problem of l ocalizing and identifying the wasted items on the moving belt is addressed. The proposed solution is based on the use of the height profile of the obj ects on the conveyor belt, sampled through a 3D optical device (laser beam plus CCD camera). With the help of fuzzy techniques, a geometrical descript ion of the items is obtained, on which basis a neural classifier can be bui lt. First experimental results are reported, confirming the validity of the presented approach. (C) 2000 Elsevier Science Ltd. All rights reserved.