Vegetation index is based on the physiological and ecological benefits of vegetation to red light and near infrared light. Scientific experiments show that plant chlorophyll needs to strongly absorb red light and blue-violet light for photosynthesis, and the absorption near the wavelength of 0.66μm is the strongest, and the absorption rate can reach 90%. The absorption intensity is related to the amount and activity of chlorophyll. The more chlorophyll, the higher its activity and the greater its absorption intensity. However, in the near infrared spectrum with the wavelength of 0.7 ~ 1. 1 micron, vegetation leaves form strong reflection, and the absorption rate is almost zero, while transmission and reflection account for almost 50% respectively. In the band of 0.35 ~ 1. 1 micron, the red light absorption peak and near infrared light reflection peak of green plants and their combinations are not found in other living things and abiotic things, so they become specific indicators for identifying vegetation, and their combinations become specific indicators for extracting vegetation information.
The data sources used by the system to extract vegetation information are Landsat-ETM, CBERS, SPOT, QUICKBIRD and MODIS. Their infrared and R-band channel numbers and wavelength ranges are shown in Table 5-2.
Table 5-2 Vegetation Index Data Source Table
Commonly used vegetation indices are:
(1) Environmental Vegetation Index (EVI): that is, the brightness difference between near-infrared band and visible red band, also known as differential vegetation index, and the expression is
EVI is sensitive to soil background; When the vegetation coverage is 15% ~ 25%, the difference increases rapidly with the increase of vegetation coverage, and when the vegetation coverage is greater than 80%, the sensitivity decreases obviously.
(2) Double Difference Vegetation Index (DDVI): that is, the brightness difference between near infrared band and visible red band minus the brightness difference between visible red band and green band (TM2), which is expressed as
The characteristic of DDVI refers to the green band sensitive to the green reflection of healthy and lush plants, which enhances the information of vegetation and can compensate the adverse effects of the atmosphere to some extent. Because green belt is sensitive to soil, it is beneficial to classify vegetation and distinguish forest types and tree species by evaluating vegetation vitality with "green peak" reflection.
(3) ratio vegetation index (RVI): that is, the ratio of near infrared band to visible red band, expressed as
RVI is sensitive to soil background. When the vegetation coverage is more than 50%, it is sensitive to the difference of vegetation coverage, but it cannot distinguish the difference of vegetation coverage less than 30%.
(4) Normalized Vegetation Index (NDVI): that is, the ratio of the difference between the near infrared band and the visible red band to the sum of these two bands, expressed as
NDVI comprehensively uses four kinds of operations, which improves the ability to distinguish soil background changes, eliminates the shadow influence of topography and community structure, and weakens the atmospheric interference, thus greatly expanding the monitoring sensitivity of vegetation coverage. It is the best index of vegetation growth and spatial distribution density, which is linearly related to plant distribution density and has good temporal and spatial adaptability, so it can also be called biomass index or standardized vegetation index.
A lot of research and comparison have been made on the processing results of the above vegetation index, and it is found that there is no significant difference in the extraction effects of various methods. However, NDVI method is the most mature and commonly used method in arid areas, because it integrates the algorithms of EVI, DVI and DDVI, and has high sensitivity to vegetation detection, wide detection range of vegetation coverage, and can eliminate shadow and radiation interference.
Based on the above principles, the automatic extraction module of vegetation coverage information organically integrates the functions of image correction, NDVI vegetation index calculation, density segmentation and so on (Figure 5-6). Users only need a few simple operations to complete the conversion from the original image data to the classification results of vegetation coverage. The core of vegetation coverage information extraction is the threshold division of density segmentation. According to many experiments and field investigation and analysis, the threshold value suitable for the classification of vegetation coverage in the working area is set in the system, which will be improved with the deepening of work and data accumulation.
Figure 5-6 Flowchart of Vegetation Coverage Information Extraction
At present, the extraction process of vegetation coverage information is mainly as follows:
(1) reflectivity inversion: reflectivity inversion is the basis of quantitative remote sensing. Regardless of the multiple scattering and cross radiation of the atmosphere, there are:
Where: DNi is the gray value of the image; ρ is the reflection of ground objects and the reflection of the outer atmosphere; GAINSi is the gain coefficient of radiation calibration, including the influence of multiplicative factors such as atmospheric transmittance and sensor wavelength response. BLASESi is the bias value of radiation calibration, including the influence of additive factors such as atmospheric radiation and sensor dark current. The gain value and offset value of radiation calibration in different bands and different bands within the same band are different. The reflectivity ρ can be inversed by the gain value and offset value of radiation calibration.
(2) Vegetation index transformation: NDVI (Vegetation Index) transformation is carried out on the image after reflectivity inversion to obtain a vegetation index image. All kinds of sampling points show dispersion matrix in NDVI images, and vegetation and non-vegetation types have great differences in NDVI images. NDVI images can distinguish vegetation from non-vegetation.
(3) Density segmentation: Vegetation index is an important index of vegetation coverage, and for remote sensing data, the reflectance in each pixel is the reflectance of canopy, not the reflectance of canopy and leaves, which can distinguish vegetation from non-vegetation, so vegetation index data is more suitable for vegetation coverage. According to the vegetation coverage, the NDVI values of 1∶ 1 10,000 vegetation index images in the main stream area and 1∶ 1 10,000 vegetation index images in key areas were divided into three levels: low coverage, medium coverage and high coverage. The classification results were retrieved by using the historical thematic data of vegetation coverage and the vegetation coverage data of ecological monitoring stations along the main stream. If the accuracy is poor, modify them.
Unsupervised classification is carried out on the vegetation index image after density segmentation, and the classification result image is obtained. The raster image data is converted into thematic graphics data with classification attributes, and combined with high-resolution fusion images, human-computer interaction interpretation is carried out, and inaccurate ground object boundaries are corrected and extracted, and finally thematic vector graphics files that meet the accuracy requirements are output.