Modification of oriented fast and rotated brief (ORB) algorithm for non-real-time images in pineapple plant disease recognition / Estacio, Mark James C.; Laurente, Christine Joy S. 6

Undergraduate Thesis: (Bachelor of Science in Computer Science) - Pamantasan ng Lungsod ng Maynila, 2024. 56

5

ABSTRACT: This study addresses challenges within the traditional Oriented FAST and Roated BRIEF (ORB) algorithm by incorporating FLANN matchers and implementing Contrast Limited Adapted Histogram Equalization (CLACHE). Our novel methodologies aim to optimize feature detection accuracy and manage image intensity variations effectively, while also substituting the Brute force matcher with FLANN to expedite and refine feature matching processes. Our approach involves refining feature detectability and integrating CLACHE to effectively handle fluctuations in image intensity, thereby enhancing adaptability to diverse environmental conditions. Furthermore, the integration of FLANN matchers streamlines feature matching, leading to accelerated processing and heightened precision in disease identification tasks. Through extensive experimentation, significant enhancements in disease identification performance are demonstrated, particularly with the increase in features from 500 to 1000. Despite encountering challenges in CLAHE integration, our findings demonstrate the effectiveness of these enhancements in surpassing existing approaches, underscoring their potential to substantially improve disease identification methodologies in agricultural settings. In summary, our study presents a comprehensive framework for enhancing the effectiveness of the ORB algorithm in disease identification, providing valuable insights for advancing computer vision applications in agriculture.

5