Pixel based classification of Lavandula sp. using high resolution UAV orthophotos

Abstract

Lavender, a member of the Lamiaceae family, is a notable plant known for its production of volatile oil. Lavender oil is prized for its antiseptic and antibiotic qualities, as well as its unique aromatic properties, making it a valuable resource in aromatherapy practices. Consequently, closely monitoring lavender plants has become a significant concern. Unmanned Aerial Vehicles (UAVs) offer a valuable solution for this task by providing high-resolution imagery through low-altitude flights and advanced digital cameras. In this study, UAVs were utilized to create orthophotos of a lavender garden. Orthophotos are meticulously corrected aerial images, ensuring consistent scale and distortion-free representation, which makes them ideal for various analytical purposes. The primary objective of the study is pixel based classification of lavender. To enhance lavender plant monitoring, machine learning techniques, specifically Support Vector Machines (SVM) and K-means clustering, were employed for binary classification. These methods were applied to analyze the orthophotos generated by the UAVs, likely to classify different sections or features within the lavender garden. In conclusion SVM provided a higher overall accuracy value for the classification results at both 1 cm and 10 cm resolutions.