Adaptive region growing for automated oil palm fruit quality recognition

Abstract

Besides rubber and rice, oil palm or Elaeis Guineensis remains as one of the most important plantation crops in Malaysia. Unfortunately, the lack of experience in oil palm fruit grading among the plucking farmers results in wrong estimation when harvesting. This affects production, negatively. Meanwhile, region growing conventional image segmentation techniques need manually or fixed initial seed selection which, actually, increases the computational cost, as well as, implementation time. Hence, the main goal of this study is to improve the seed region growing algorithm in order to gain higher accuracy in segmenting color information for oil palm fruit image. This study presents n-Seed Region Growing (n-SRG) for color image segmentation by choosing adaptive numbers of seed. The data sample consists of 80 images which comprises and two ripeness classes (ripe and unripe).The proposed work has out-performed the k-mean clustering method with 86% and 80% of average accuracy rates correspondingly. © 2013 Springer International Publishing.