Syerlin Aprilia Delvi, Agung Ramadhanu

A Implementasi K-Means Clustering dalam Segmentasi Citra Hewan pada Kucing, Kambing, dan Burung

Introduction

A implementasi k-means clustering dalam segmentasi citra hewan pada kucing, kambing, dan burung. Implementasi K-Means Clustering untuk segmentasi citra hewan (kucing, kambing, burung) dalam pemrosesan citra digital. Evaluasi efektivitas pemisahan objek dari latar belakang secara otomatis.

Abstract

Image segmentation is one of the most important challenges in digital image processing because it determines the successof separating the main object from the background so that visual information can be further analyzed. The problem ariseswhen the object has complex color, texture, and shape characteristics, as in animal images that often have color patternssimilar to their surroundings, making object boundaries difficult to distinguish clearly. This study aims to apply the KMeans Clustering method in the process of animal image segmentation—specifically for cats, goats, and birds—and toevaluate its effectiveness in identifying and separating the main object from the background. The method used is the KMeans Clustering algorithm, an unsupervised learning technique that groups image pixels based on color similarity in theRGB color space through an iterative process until centroid stability is achieved and clusters representing different imageregions are formed. The results show that the K-Means method can produce good segmentation performance for imageswith uniform lighting and simple backgrounds but experiences a decrease in accuracy when the object’s color is similar toits environment. Overall, this algorithm is effective, simple, and can serve as a foundation for developing automatedanimal image identification and classification systems

Review

The paper, "A Implementasi K-Means Clustering dalam Segmentasi Citra Hewan pada Kucing, Kambing, dan Burung," addresses a significant challenge in digital image processing: effective image segmentation, particularly for animal subjects where complex characteristics and environmental similarities can obscure object boundaries. The authors propose applying the K-Means Clustering method to segment images of cats, goats, and birds, aiming to evaluate its efficacy in separating foreground objects from their backgrounds. This study is timely and relevant, as accurate segmentation is a crucial prerequisite for subsequent image analysis, object identification, and the development of automated classification systems. The methodology centers on the K-Means Clustering algorithm, an unsupervised technique that groups pixels based on color similarity within the RGB space, iterating until centroid stability is achieved. While K-Means is valued for its simplicity and computational efficiency, the abstract commendably highlights both its successes and inherent limitations within this application. The method reportedly yields good segmentation performance under ideal conditions, such as uniform lighting and simple backgrounds. However, a critical drawback emerges when the object's color closely matches its environment, leading to a notable decrease in accuracy—a common real-world challenge for animal imagery. The findings suggest that K-Means serves as an effective and simple foundational algorithm for animal image segmentation, especially in controlled environments. Nevertheless, its acknowledged struggle with color similarity between the object and background indicates that its standalone application may be insufficient for scenarios involving highly complex or camouflaged animals. For the development of truly robust automated animal identification and classification systems, future research should explore integrating additional features beyond color, such as texture and shape descriptors, or investigate hybrid clustering approaches. This would likely enhance the algorithm's resilience and accuracy across a more diverse and challenging range of animal images.

Full Text

Comments

You need to be logged in to post a comment.