Extracting Lotus Fields Using the Spectral Characteristics of GF-1 Satellite Data

Abstract

The lotus (Nelumbo nucifera Gaertn.) is an aquatic plant that grows in shallow water and has long been cultivated in South China. It can improve the incomes of farmers and plays an important role in alleviating poverty in rural China. However, a modern method is required to accurately estimate the area of lotus fields. Lotus has spectral characteristics similar to those of rice, grassland, and shrubs. The features surrounding areas where it is grown are complex, small, and fragmented. Few studies have examined the remote sensing extraction of lotus fields, and automatic extraction and mapping are still challenging methods. Here, we compared the spectral characteristics of lotus fields and other ground objects and devised a remote sensing method for the rapid extraction of lotus fields. Using this method, the extraction accuracy of lotus was 96.3%. The Kappa coefficient was 0.926, which is higher than those of the unsupervised K-means classification, Mahalanobis distance, and support vector machine supervised classification, and demonstrates the potential of this method for extracting and mapping lotus fields by remote sensing.