Comparison of automatic extraction methods of vegetation cover based on the grassland positioning observation station

Vegetation cover is an important index in grassland ecological monitoring. In this study, positioning observation stations were used to obtain the image of the sampling plot, and vegetation cover extraction algorithms suitable for different grassland positioning observation stations were selected and compared. This was conducted with the aim of solving the problem of how to automatically extract vegetation cover indices from images of different plots in continuous positioning observation. Different image segmentation methods such as the greenness index method, random forest (RF), support vector machine (SVM), and back propagation (BP) neural networks were used to obtain the cover extraction results for different grassland types. The pros and cons of threshold segmentation methods such as the greenness index and machine learning methods, the reasons for the different classification effects of three machine learning algorithms, and the main reasons for the error of coverage value were discussed in this paper. The results have shown that the machine learning algorithm could be flexibly applied to the rapid automatic extraction of vegetation cover in the quadrat image of the positioning observation station. The greenness index threshold segmentation method applied to the segmentation of vegetation cover was relatively poor. RF algorithm has a higher level of accuracy in the segmentation of alpine grassland. SVM has a higher level of accuracy in temperate grassland and temperate desert grassland. The BP neural network has more advantages in the cover extraction of alpine meadow. This study can provide an important reference for the development of information and intelligent monitoring equipment for grassland ecological monitoring in the new era.