Segmentation of Arecanut Bunches Using Deep Learning Technique

Abstract

Image segmentation plays a crucial role in computer vision, enabling the extraction of detailed features from images for a wide range of applications. Convolutional Neural Networks (CNNs) have demonstrated exceptional performance in achieving accurate segmentation outcomes across various domains. This study presents a novel method for segmenting unharvested arecanut bunch images using CNN models. An optimized U-Net model was utilized for segmentation, and a comparative analysis was conducted across three different color spaces: RGB, saturation, and grayscale. The results indicate that the proposed model performs best with RGB images, achieving a Dice coefficient of 91.15%, which is notably higher compared to the segmentation of images in the other two color spaces. This research underscores the superior accuracy of using RGB images for the segmentation of arecanut bunches, offering valuable insights for applications in precision agriculture.