Classification Of Wheat Crop Using Remotely Sensed Multi-spectral Planet-scope Temporal Data


Agriculture plays a vital role in the economies of developing countries and provide the main source of food, income and employment for general public. Monitoring and assessment of the crop yield is a crucial task and is critical in ensuring good agricultural management. We propose the monitoring and classification of wheat crops through remote sensing by utilizing satellite imagery. In our research work, we have utilized multi-spectral imagery of Planet-Scope satellite for the classification of wheat crop. The imagery used is a temporal stack of remotely sensed imageries obtained on various dates with reference to the phenological cycle of wheat. We employ three different machine learning classifiers i.e., Artificial Neural Network (ANN), Support Vector Machine (SVM) and Minimum Distance (MD) classifier for the wheat crop classification. Confusion matrix and Kappa Coefficient (Kappa Coefficient) analyzes the performance of these three classifiers. The results obtained shows that ANN with an overall accuracy of 98:7031%and Kappa Coefficient equivalent to 0:9825 outperforms the SVM and MD classifiers having the overall accuracy of 85:2005% and 73:1604% and Kappa Coefficient values of 0:8097 and 0:6455, respectively.