Identification of fruit and branch in natural scenes for citrus harvesting robot using machine vision and support vector machine

• DOI: 10.25165/ijabe.v7i2.1215
• Publication Date: 2014-04-01
• Authors: Qiang Lü; Cai JianRong; Bin Liu; Lie Deng; Yajing Zhang
• Keywords: Machine Vision; Identification

Abstract: With the decrease of agricultural labor and increase production cost, researches on citrus harvesting robot (CHR) have received more attention in recent years. For success robotic safety robot, identification mature fruit obstacle is priority harvesting. In this work, a machine vision system, which consisted color CCD camera computer, was developed to achieve these tasks. Images trees were captured under sunny cloudy conditions. Due varying degrees lightness position randomness fruits branches, red, green, blue values objects images are changed dramatically. The traditional threshold segmentation not efficient solve problems. Multi-class support vector (SVM), succeeds by morphological operation, used simultaneously segment branches study. recognition rate 92.4%, branch diameter than 5 pixels, could be recognized. results showed that algorithm detect for CHR. Keywords: citrus, vision, (CHR), branch, identification, multi-class (SVM) DOI: 10.3965/j.ijabe.20140702.014 Citation: Lu Q, Cai J R, Liu B, Deng L, Zhang Y J. Identification natural scenes using machine. Int Agric & Biol Eng, 2014; 7(2): 115－121.