Automatic image segmentation of oil palm fruits by applying the contour-based approach

Abstract

Nowadays, computer vision-based automation systems in the field of oil palm plantations are still being developed. The ripeness grading for quality control of oil palm fruit and the fruit picking robots are the implementation examples of those systems. Such a system requires a pre-processing called segmentation. Segmentation is a complex process and will be more challenging if the image of oil palm fruit is captured in the natural background moreover with uneven brightness. In this work, the segmentation method of oil palm fruit is proposed by applying the contour-based approach due to the oil palm fruit that has various shapes and colors. This method implemented the Canny algorithm combining with several operations of morphology and reconstruction to remove the noise. The proposed method is evaluated using a local dataset consisting of 160 images with three kinds of ripeness grading (60 raw, 50 under-ripe, and 50 ripe). Overall, the average of segmentation accuracy achieves of 90.13% with the error rates of false positive and false negative are 2.92% and 5.20%, respectively.