Object-oriented Vegetation Classification Method based on UAV and Satellite Image Fusion
Fuchao Zhao, Xiaoming Wu, Shuai Wang
Abstract
Nowadays, space remote sensing technology has become one of the most important means for people to obtain geographic information and environmental information. Satellite multi-spectral data contains rich spectral information in multiple bands. The method of remote sensing monitoring of ground vegetation classification and identification represented by satellite is widely used. Because the Unmanned Aerial Vehicle (UAV) has obvious advantages such as small size, strong timeliness, flexible operation and low cost, it is widely used as a remote sensing platform for disaster monitoring, environmental detection, vegetation distribution information monitoring, etc. And it quickly becomes an important way to get the vegetation type of the study area. Compared with satellite multi-spectral images, UAV images have high resolution and rich spatial information, but lack spectral information for vegetation identification. So this article is based on satellite images, in order to improve the precision of ground vegetation in the area of classification and recognition, the UAV images and satellite multi-spectral images data pixel level fusion, combined with object-oriented supervised classification method, the integration of data with random forests (RF), support vector machine (SVM) and maximum likelihood estimation (MLE) method for vegetation classification identification precision and validation. Compared with the data before fusion, the statistical results of the confusion matrix output show that the accuracy of vegetation classification recognition under the combination of object-oriented supervised classification method has been significantly improved.