Segmentation of Stacked Leaf Images for Enhanced Visual Monitoring Using Gaussian Mixture Models (GMM) Algorithms

Piles of leaves that are blocked by other objects hinder visual monitoring of the condition of the leaves. Therefore, this research aims to identify and separate images of stacked leaves from the background. Advances in artificial intelligence technology with machine learning algorithms make it possible to research image segmentation of stacked leaves. This research succeeded in conducting experiments by segmenting stacked leaf images using Machine Learning algorithms, namely K-Means Clustering, Fuzzy C-Mean (FCM), and Gaussian Mixture Models (GMM). This algorithm was evaluated using Mean Squared Error (MSE) and peak signal-to-noise ratio (PSNR). The experimental results show that the GMM algorithm achieves the lowest MSE value of 44.28781, outperforming both the K-Means and FCM algorithms. Likewise, the PSNR value on GMM tends to be greater than the PNSR value on K-Means and FCM, namely 31.66796dB. The experimental results prove that in the case of stacked leaf image data, GMM appears to be the best choice compared to the two algorithms tested in this research.