A Vision System to Efficiently Harvest Fruits Which Grow in Clusters

Abstract

Automatic fruit harvesting systems usually come up against some difficulties because of the environment in which they operate. In this paper, we will consider the problem of distinguishing individual fruits which grow in bunches or clusters. A cluster may contain fruits with similar color and texture, what makes them appear in the image as a unique uniform region. Therefore, it is not easy to distinguish the individual fruits using traditional detection methods. We propose a new vision system composed of two stereoscopic cameras and a matrix of laser diodes that is able to distinguish fruits that grow in bunches. A prototype of a harvesting robot based on it has been developed and tested with real strawberry crops in hydroponic greenhouses. Some examples of these experiments are included.

Primena automatskih sistema za ubiranje voća obično nailazi na teškoće obziorm na okolinu u kojoj rade. U ovom radu razmatran je problem identifikacije pojedinačnih plodova koji se razvijaju u grupi plodova. Skupovi plodova mogu da sadrže u sebi plodove slične boje i teksture što dovodi do toga da se na slici pojave kao uniformna jedinstvena oblast. Iz toga razloga nije jednostovano identifikovati pojedinačne plodove upotrebom uobičajenih metodama. U radu je dat predlog novog vizuelnog sistema koga čini dve steroskopske kamere i matrične laserske diode koje mogu da razdvoje plodove koji rastu u skupu. Na osnovu ovog sistema napravljen je prototip robota za ubiranje koji je ispitan u ubiranju jagode gajene u zaštićenom prostoru, u hidroponskom sistemu. Prikazani su neki primeri iz ovog eksperimenta.