Image quality analysis
Among the 128 band HyMap data obtained, four bands (437φ, 88 1φ, 16 18φ, 163 1φ) failed due to instrument reasons, and no data were collected. Available data is actually only 124 band data. However, there are basically no characteristic spectra of minerals in these four bad zones, which has little effect on the extraction of mineralization and alteration information and mineral mapping.
Through visual inspection and statistical analysis, it is found that the image data quality of 124 band is good. Among them, 1396φ (Figure 4-2-2(a)) and141φ (Figure 4-2-2(b)) are located at the absorption wavelength of the atmospheric water belt, which basically reflects the characteristics of atmospheric water vapor, and the images of ground objects are almost invisible or blurred. These two bands can be used to extract atmospheric water vapor content and provide basis for atmospheric correction. The rest of the 122 band images are very clear, as shown in figure 4-2-2(c).
Signal-to-noise ratio analysis of 4.2.2.2 data
Signal-to-noise ratio is one of the most important indexes in imaging spectrum. Because the imaging spectrum mainly distinguishes and identifies ground objects according to the subtle characteristics of the reconstructed spectrum, the signal-to-noise ratio directly affects the extraction and application effect of spectral characteristics of rock and mineral geological bodies. There are many methods to analyze the signal-to-noise ratio: theoretical calculation, simulation analysis and statistical analysis. The signal-to-noise ratio analysis of HyMap data adopts statistical analysis method. Statistical analysis only uses 122 band data (excluding the number of bad bands and bands with extremely low noise). According to the original data of this 122 band and the reflectivity data generated by ground synchronous calibration, the signal-to-noise ratio is counted, and the selected yellow soil is statistically analyzed. The surface is evenly distributed with the range of 10× 10 pixels. Figure 4-2-3 is the signal-to-noise ratio curve of HyMap data in visible light, near infrared and short-wave infrared bands. As can be seen from the figure, the signal-to-noise ratio of the original data is between 80 ~ 140, the average signal-to-noise ratio in visible light is around 120, and the signal-to-noise ratio in short-wave infrared is lower, between 70 ~ 100, with an average of 85. This can also be explained by the fact that the dark current in the infrared range of each band (Figure 4-2-5(b)) is nearly 6 times higher than that in the visible band (Figure 4-2-5(a)). At the wavelength of 400 ~ 2500 φ, the signal-to-noise ratio curve (Ref) of the corrected reflectivity data is higher than that of the original data (DN), especially at the wavelength of visible light, which is 40 ~ 60 φ, and the two curves are basically the same under short-wave infrared.
Table 4-2- 1 HyMap band and wavelength calibration parameter table of imaging spectrometer
sequential
Figure 4-2-2 Comparison of HyMap Image Quality in Experimental Area
Stability Analysis of 4.2.2.3 Data
Statistical analysis of the maximum, minimum, average and variance of spectral irradiance response values of known calibration lamps can reflect the stability of instrument response data. Figure 4-2-4 shows the maximum, minimum, average and variance curves of the spectral response data of HyMap calibration lamp in flight band. As can be seen from the figure, the curves of the maximum, minimum and standard deviation of the spectral response value of the calibrated light almost coincide with the average curve, which shows that the spectral information data of the ground objects obtained by the imaging spectrometer flying online has little drift and is stable and reliable. The data value of noise (dark current) curve of on-line flight detection is small and the deviation is very small (Figure 4-2-5). Combined with the characteristics of 16 bit (0.65536), such as large dynamic range of data quantization value and no signal saturation when the surface reflectivity is maximum. It also shows that the obtained data is stable and reliable.
Figure 4-2-3 Signal-to-Noise Ratio Curve of HyMap Imaging Spectral Data in Experimental Area
Figure 4-2-4 Corresponding Curve of Hymap Online Flight Alignment Calibration Lamp
The top-down arrangement order of curves in the figure corresponds to the top-down interpretation order of words on the right.
MNF analysis of 4.2.2.4 data
Figure 4-2-5 Dark Current Curve of Hymap at 480nm and 2200nm Wavelengths
Minimum noise separation method MNF(mini? μm noise fraction) is an orthogonal transformation, similar to principal component analysis (PCA) of multispectral remote sensing data. Orthogonal transformation can separate signal from noise, and the eigenvalues of the first few frequency bands after transformation are much larger than those of the last few frequency bands. Moreover, the images of ground objects displayed in these bands are clear and concentrate most of the spectral information of ground objects. The spectral information of ground objects contained in the later bands decreases in turn, while the noise increases (Green, A.A., Berman, M., Switzer, P. et al., 1988). Lee J.B et al.,1990; Yang Kai, 2003). Figure 4-2-6 is the MNF orthogonal transform band image of 27 bands of short-wave infrared in the seventh band of the first area: MNF band 1 represents the brightness background of the whole band, that is, the spectral background, which is brighter than other MNF bands in the image; The spectral information of rock and mineral geological bodies is concentrated in 2 ~ 6 bands, and the image is very clear, but noise also appears gradually; The texture of the seventh zone, such as spatial topography, is very clear; The system noise obviously appears in the 7th to 9th frequency bands. After 10 band, the random noise is very strong, which almost covers the spectral information and spatial information of geological bodies. The visible and near infrared bands of this band are analyzed by similar methods, and the results show that the signal and noise of the data can be separated. Therefore, HyMap data signal and noise distribution are normal, and the data quality in visible light, near infrared and short-wave infrared is good.
Fig. 4-2-6 MNF transform for separating band spectrum signal from noise
Wavelength detection of 4.2.2.5 characteristic spectrum
Remote sensing of rock and mineral geological bodies by imaging spectrum is to directly match and identify the spectral characteristics of standard rocks and minerals by using the spectral information of rock and mineral characteristics extracted from imaging spectrum data. Whether the wavelength position of the identified characteristic spectrum is correct or not will affect the effect of rock and mineral identification and mapping. Therefore, it is necessary to detect the wavelength of quasi-reflectivity spectral characteristic data retrieved from imaging spectral data. The detection method is to compare the absorption characteristic wavelength position of HyMap reflection spectrum of a known point with the spectral characteristics of ground objects at the same point measured by a spectrometer with higher spectral resolution. Fig. 4-2-7(a) is the characteristic spectral curve of sericite-containing endmember minerals extracted from HyMap imaging spectral data in 2000 ~ 2500 φ band. The Al-OH groups in these three sericite minerals are located at 2220φ (upper part), 22 10φ (middle part) and 2 195φ (lower part) respectively. Fig. 4-2-7(b) The short-wave infrared spectrum characteristic curve of sericite samples measured by PIMA ground spectrometer with higher spectral resolution at the same geographical location shows that the Al-OH groups in sericite samples have obvious absorption around 22 18φ (top), 2206φ (middle) and 2 194φ (bottom) respectively. The wavelength differences of these three sericite minerals containing Al-OH in ground and air are 2φ, 4φ and 1φ respectively.
Fig. 4-2-7 Comparison of characteristic wavelengths between HyMap image spectrum and PIMA spectrum of three sericite-containing mineral samples
Fig. 4-2-8(b) is the characteristic spectrum of ion-containing calcite minerals measured from aerial HyMap imaging spectrometer and ground PIMA-II at the same geographical location (fig. 4-2-8(a)). From these two spectral characteristic curves, the reflectivity curve of aviation is slightly higher than that of ground test, and there is a strong characteristic absorption valley near 2338φ, which basically overlaps. Therefore, it is considered that the maximum wavelength shift of spectral characteristics of rock and mineral geological bodies extracted from HyMap may be 1 ~ 2φ.
4.2.2.6 abstract
Through the above-mentioned visual analysis, data signal-to-noise ratio analysis, data stability analysis, MNF orthogonal transformation of data, wavelength detection analysis and flight quality analysis, it is considered that the quality of HyMap imaging spectral data obtained in East Tianshan experimental area is as follows: ① Only four bands 128 of spectral image data obtained by visible light, near infrared and short-wave infrared cannot be used; ② The signal-to-noise ratio of visible light, near infrared and short-wave infrared data decreases in turn, and the visible light is about 120, the near infrared is about 1 10, the short-wave infrared is about 1000 ~ 1800φ, and the short-wave infrared is about1. (3) According to MNF transform analysis, the signal and noise distribution of HyMap data is normal; (4) The data of the calibration lamp is stable, the dynamic range of the data is large, and the data signal of high reflectivity rocks and minerals is not saturated; ⑤ The maximum possible wavelength shift range for extracting characteristic spectra of typical rocks and minerals is 1 ~ 2φ.
To sum up, the HyMap imaging spectral data obtained in this experiment has high signal-to-noise ratio, clear images of rock and mineral geological bodies and moderate contrast. The extracted spectral characteristic curve of rocks and minerals is close to the measured spectral curve of rocks and minerals. The spectral information of the data is reliable and can be used for rock and mineral mapping.
Fig. 4-2-8 Comparison of Characteristic Spectra of Calcite-bearing Minerals Obtained from Aerial HyMap and Ground PIMA-Ⅱ