As one of the most advanced subjects in remote sensing field, hyperspectral plays an important role in Earth observation. Hyperspectral has the characteristic of " combination of image and spectrum " , which contains hundreds of bands and contains abundance spectral information. In recent years, with the development of airborne hyperspectral technology, it has achieved many results in precision agriculture, geological mapping, ecological environment surveys and so on. However, due to the development of payload technology, it is difficult for traditional hyperspectral data to have high spatial resolution as well as high spectral resolution without affecting the efficiency of data acquisition. Due to the relative lack of spatial information, hyperspectral data have been paid more attention to the spectrum itself, and the significance of spatial-spectral correlation information is often ignored. In terms of applications, there are few cases of hyperspectral data in urban remote sensing applications and local scale ecological environmental monitoring. Based on the data obtained from the high-resolution airborne hyperspectral loading developed recently in our country, the present situation of the riparian eco-environment in the Liaohe River region of the Liaozhong was analyzed. Using multi-scale fully convolutional neural network to extract the feature information of the data space spectrum. Based on the classification of water, crops, forest, grass and buildings in the demonstration area, the degradation of riparian zone caused by human activities in the demonstration area was analyzed. Green Index, Light Utilization Index and leaf pigment index were inversed by hyperspectral method, and the health status of forest and grass in the demonstration area was classified. The results show that the high resolution hyperspectral data can not only describe the spatial distribution of the objects, but also retrieve the health status of the objects, which is of great significance to the refinement of the ecological Environmental monitoring in local areas.
|