Coffee plantation area recognition in satellite images using Fourier transform

• Main Authors: Wan-Ling Chen, 陳婉菱
• Other Authors: Du-Ming Tsai
• Format: Others
• Language: zh-TW
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/30362776295443733616

碩士 === 元智大學 === 工業工程與管理學系 === 104 === In this study, a machine vision method is proposed for coffee plantation area recognition in satellite images. It automatically segments the coffee fields with row-planting from tree forest and irrelevant regions in the image. The result can be used for coffee yield estimation to improve the supply and demand of coffee production. It can thus reduce the fluctuation of coffee commodity prices in the world market. Commercial coffee plantation grows coffee trees in row along specific direction to increase production yield and ease farming management. The row-planting pattern of coffee trees shows structural texture in the satellite image. This study presents a Fourier transform-based method to extract structural features from the power spectrum for image segmentation. The coffee plants and tree forest present the same color tone in the image; thus, color cannot be used as the discriminant feature. The row-planting of coffee in a small image patch shows a highly structural texture in the spectrum, and the main frequency of the spectrum represents the number of rows in the patch. In contrast, a natural tree forest shows no structural pattern in the spectrum because trees grow in random directions. The Fourier transform can be effectively utilized to extract structural features for robust segmentation. This study also analyzes the color indicators of RGB, HSI, Lab and Lhs color models, and evaluates their performance on the segmentation of green plant regions in the image. The color segmentation result can reduce the computation load of the Fourier transform in an image. This study analyzes satellite images of coffee plantation areas at different times, weather conditions (lighting) and planting modes (parallel and trellis planting) in Latin America, Africa, the Middle East and South Asia, and East Asia and the Pacific Islands. The experimental results have shown that the Fourier-based structural features can provide the correct segmentation results to distinguish planting coffee trees and tree forest in different environments.