Flower species identification and coverage estimation based on hyperspectral remote sensing data

Monitoring grass species and coverage accurately makes a significant contribution to species diversity research and sustainable development of grassland ecosystem. Plants grown in grassland usually own unique spectral characteristics in florescence. Compared with the nutrient stage, species are more easily identified during florescence. In this study, flowers such as Galium verum Linn., Hemerocallis citrina Baroni, Serratula centauroides Linn., Clematis hexapetala Pall., Lilium concolor var. pulchellum, Lilium pumilum and Artemisia frigida Willd. Sp. PI. were identified, using some canopies spectra analysis and feature extraction methods. Validation shows that when the coverage of flowers is greater than 10%, the accuracy of identification methods will be higher than 90%. Based on this result, linear unmixing model is adopted to calculate the area ratios of flowers in quadrates. Results show that linear unmixing model is an effective method for estimating the coverage of grassland flowers with the mean retrieval error of about 4%.