Using Hyperspectral Data to Identify Crops in a Cultivated Agricultural Landscape - A Case Study of Taita Hills, Kenya

Abstract

Recent advances in hyperspectral remote sensing techniques and technologies allow us to more accurately identify larger range of crop species fromairborne measurements. This study employs hyperspectral AISAEagle VNIRimagery acquired with9 nmspectral and 0.6 m spatial resolutions over a spectral range of 400 nm to 1000 nm. The area of study is the Taita hills in Kenya. Various crops are grown in this region basically for food and as an economic activity.The crops addressed are: maize, bananas, avocados, and sugarcane and mango trees. The main objectives of this study were to study what crop species can be distinguished from the cultivated population crops in the agricultural landscape and what feature space discriminates most effectivelythe spectral signatures of different species. Spectral Angle Mapper (SAM)algorithm together with some dissimilarityconcepts was applied in this work. The spectral signatures for crops were collected using accurate field plot maps. Accuracy assessment was done using independent training vector data. We achieved an overall accuracy of 77% witha kappa value of 0.67. Various crops in different locations were identified and shown.