Hebei Normal University uses SOC710 for grassland degradation identification
College of Resources and Environmental Sciences, Hebei Normal University
Hebei Province Environmental Evolution and Ecological Construction Laboratory
Hebei Normal University Tourism Department
Environmental Protection Nanjing Institute of Environmental Sciences
Grassland ecosystems play an important role in the development of animal husbandry, maintaining soil and water, and maintaining ecological balance. It is of great significance to monitor grassland degradation in real time and accurately. Hyperspectral remote sensing can greatly improve the recognition accuracy of vegetation structure degradation in grassland degradation process, and open up new fields for grassland degradation research. The selection and extraction of characteristic bands is essential when using hyperspectral remote sensing technology to identify vegetation structure degradation.
The selected observation objects are degraded indicator species and main dominant species in the Bashang area of ​​Hebei Province. Through field investigation, the selected indicator species are Stellera chamaejasme, Artemisia frigida, and Star sylvestris. The main dominant species are Carex and Leymus. These two plants belong to the extensive plant in the study area.
The selected instrument is a SOC710VP portable visible/near-infrared hyperspectral imager developed by American SOC. The spectrum of the instrument is from 400 to 1000 nm from visible to near-infrared, with a spectral resolution of 4.68 nm and SOC710. The supporting software SRAnap710 pre-processing software performs data calibration processing and uses ENVI software for data processing and analysis.
Feature band selection is to select feature subsets from the original set, which can effectively describe the feature information of the spectrum and achieve dimensionality reduction. The value of the band i is 373-1033 nm, and the spectral reflectance mean confidence interval of all bands constitutes the mean confidence interval band of the spectral reflectance. The principle of extracting the characteristic band based on the spectral reflectance mean confidence interval is shown in Fig. 1.
It can be seen from Fig. 1 that in the band intervals [a, b] and [c, d], the spectral reflectance mean confidence interval bands of the two types of vegetation overlap, which is not suitable for vegetation identification. By eliminating the overlapping parts of the reflectance mean confidence interval, the optimal vegetation identification band is selected to achieve the purpose of dimension reduction and the selection of feature bands.
The comparison of the original spectral curves of the 5 plantings based on the characteristic band of the mean confidence interval is shown in Figure 2-4.
It can be seen from Fig. 2-4 that the characteristic band of Stellera chamaejasme is 402-412 nm; the characteristic bands of Artemisia frigida are 627-689, 715-929 and 929-1033 nm; and the characteristic band of Starhair is 705-721 nm. It can be seen that the characteristic band of Stellera chamaejasme is located in the range of visible light spectrum, which may be related to the height of the Stellera chamaejasme plant, the large leaf area and the high chlorophyll content. The characteristic band of Potentilla angustifolia is located at the “red edge†position, which may be related to vegetation growth period and vegetation growth. The situation is related; the overall trend of the reflectivity of Artemisia frigida is much larger than that of Carex and Leymus, and its characteristic band is involved in the visible light band, the "red edge" position and the near-infrared band, and may be accompanied by fluff on the surface of Artemisia frigida. It reduces the reflection characteristics of chlorophyll and is therefore easily related to other vegetation.
The logarithmic spectral curve of the 5 plantings based on the logarithmic spectral curve degradation indicating the characteristic band of the plant is shown in Figure 5-7.
It can be seen from Fig. 5-7 that the content and depth of Stellera chamaejasme, Artemisia frigida, and Potentilla angustifolia are compared with Carex and Leymus, respectively, the reflection peak at 550 nm, the blue valley at 420 nm, and the red valley at 670 nm. Increasingly, the difference characteristics of the reflectance spectral curves between species are more obvious, which is more conducive to the identification of characteristic bands. The logarithmic spectral curves of the degraded indicator species and the dominant species are compared with the original spectral curves, and the extraction results of the characteristic bands are basically the same. The logarithmic spectral analysis based on the mean confidence interval showed that the characteristic band of Stellera chamaejasme was 402-412nm, and the characteristic bands of Artemisia frigida were 611-689, 758-924 and 940-1038nm. 705-721nm.
The Manhattan distance method was used to evaluate the recognition effect of the screening characteristic bands on the degraded indicator species and the dominant species. The degraded indicator species Stellera chamaejasme L., Artemisia scopariae and S. sylvestris were combined with the main dominant species of Carex and Leymus chinensis. For the group, the test distance was used to test the Manhattan distance. The results are shown in Table 1.
Comparing the Manhattan distance value of the same planting with the Manhattan distance value of the different plantings, it is found that no matter which kind of vegetation is the experimental sample or the test sample, the Manhattan distance value of the planting is significantly smaller than the Manhattan distance value of the different planting. The Manhattan distance of Artemisia frigida and Carex, the Manhattan distance of Artemisia frigida and Leymus chinensis, and the Manhattan distance of Artemisia frigida and Artemisia scoparia are larger than those of Stellera chamaejasme and Potentilla chinensis and Carex and Leymus chinensis. It indicates that the types of vegetation of Artemisia frigida and Carex and Leymus chinensis are different, which is easy to distinguish and identify.
Ningbo DSS Intelligent Technology Co., Ltd , https://www.dssking.com