Leveraging Region Growing Algorithm for Fruit Identification and Extraction in Natural Scene Images

This paper presents an efficient method for fruit identification and extraction from natural scene images by leveraging the Region Growing Algorithm. The method effectively identifies and segments fruits from complex and cluttered backgrounds, which poses a significant challenge in image analysis tasks in agricultural environments. The proposed technique employs a region growing approach that starts with a seed pixel and iteratively adds neighbouring pixels based on similarity criteria, forming a region corresponding to a fruit. The algorithm is designed to function well under varying lighting conditions, with different fruit types, and amidst complex background clutter. Experimental results demonstrate that this approach effectively segments fruits, achieving high accuracy in real-world image conditions. The method is robust and insensitive to noise, blur, variations in fruit size and shape, colour differences, uneven surface textures, and changing lighting conditions. The proposed method achieved an accuracy of 95% to 98%.