Abstract
The accuracy of supervised image classification is highly dependent upon several factors such as the design of training set (sample selection, composition, purity and size), resolution of input imagery and landscape heterogeneity. The design of training set is still a challenging issue since the sensitivity of classifier algorithm at learning stage is different for the same dataset. In this paper, the classification of RapidEye imagery with balanced and imbalanced training data for mapping the crop types was addressed. Classification with imbalanced training data may result in low accuracy in some scenarios. Support Vector Machines (SVM), Maximum Likelihood (ML) and Artificial Neural Network (ANN) classifications were implemented here to classify the data. For evaluating the influence of the balanced and imbalanced training data on image classification algorithms, three different training datasets were created. Two different balanced datasets which have 70 and 100 pixels for each class of interest and one imbalanced dataset in which each class has different number of pixels were used in classification stage. Results demonstrate that ML and NN classifications are affected by imbalanced training data in resulting a reduction in accuracy (from 90.94% to 85.94% for ML and from 91.56% to 88.44% for NN) while SVM is not affected significantly (from 94.38% to 94.69%) and slightly improved. Our results highlighted that SVM is proven to be a very robust, consistent and effective classifier as it can perform very well under balanced and imbalanced training data situations. Furthermore, the training stage should be precisely and carefully designed for the need of adopted classifier.
Highlights
With the increasing number of earth observation satellites and recent advancement in remote sensing data analysis, there has been a tremendous increase in earth observation applications ranging from environmental monitoring and mapping, climate dynamics up to the disaster monitoring and risk assessment (Khatami et al, 2016).Remote sensing provides crucial and efficient information about the earth’s land cover in local or global scales accurately and temporally these information is important for policy and decision makers on many socioeconomic and environmental issues (Townshend, 1992; Kavzoglu, 2009)
This paper aims to evaluate the how much the classifiers have been influenced with imbalanced training data for mapping the crop types and comparative performance for each classifiers with RapidEye imagery
Experimental results suggest that Maximum Likelihood (ML) and NN classifications were negatively affected by imbalanced training data in resulting a reduction in accuracy while Support Vector Machines (SVM) is not affected significantly and slightly improved
Summary
With the increasing number of earth observation satellites and recent advancement in remote sensing data analysis, there has been a tremendous increase in earth observation applications ranging from environmental monitoring and mapping, climate dynamics up to the disaster monitoring and risk assessment (Khatami et al, 2016).Remote sensing provides crucial and efficient information about the earth’s land cover in local or global scales accurately and temporally these information is important for policy and decision makers on many socioeconomic and environmental issues (Townshend, 1992; Kavzoglu, 2009).Due to the rapid population growth and global climate change, the sustainable management of agricultural as well as natural resources are becoming crucial for countries regarding to increasing necessity of food and water (Forkuor et al, 2014; Kim and Yeom, 2015). The number of earth observation satellites incorporating the sensor sensitive to chlorophyll content of vegetation as well as its related environmental applications have been increasing over the last few years (Omer et al, 2015; Gärtner et al, 2016). As an such example of recently launched satellite Sentinel-2A (2015) which is a European high resolution and multispectral imaging system offers 13-multispectral bands with spatial resolutions of 10,20 and 60 meters including three different red-edge and one nearinfrared bands as those are useful for agricultural, ecological, and forestry applications (Immitzer et al, 2016). The contribution of red-edge band over the agricultural areas for the classification as well as on feature extraction have been tested and proved in many studies (Schuster et al, 2012; Adelabu et al, 2014)
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
