Rock and mineral identification method based on mathematical theory and mineral spectrum (multispectral or hyperspectral) characteristics (Zhou Chenghu et al.,1999; Gan Puping, 2002), such as image enhancement, image transformation, spectral angle identification, optimal density segmentation, cross-correlation matching identification, imaging spectral rock and mineral identification based on complete spectral model, automatic mineral identification using neural network, etc., although some achievements can be achieved in the exposed experimental area, it seems that there is no prospect of popularization and application in the vegetation-covered area. The main reasons are the mixing of pixels and the inherent uncertainty of rock spectrum (Cheng Cheng et al., 2004). Lithology remote sensing is difficult, and the fusion of remote sensing information with geophysical and geochemical information, especially airborne radioactive information, may be one of the development trends of remote sensing lithology identification research.

The content and distribution characteristics of radioactive elements in different types of rocks are different (Department of Geochemistry, Wuhan Institute of Geology,1979; Liu Yingjun et al., 1984), Table 8. 1 lists the contents and ratios of major natural radioactive elements in several major rocks. As can be seen from the table, the content of radioactive elements in igneous rocks increases with the increase of rock acidity; The content of radioactive elements in fine clastic rocks is higher than that in coarse clastic rocks. Among metamorphic rocks, the content of radioactive elements in low metamorphic rocks is higher than that in deep metamorphic rocks. For granite, the content of radioactive elements is related to the geological age of the rock, that is, the older the age, the higher the content of radioactive elements. It can be seen that the radioactive characteristics in rocks can provide a lot of information to distinguish lithology from strata. Moreover, when rocks undergo epigenetic changes, some radioactive elements (such as thorium) or the ratio of radioactive elements are relatively stable, and vegetation has little effect on rock radioactivity measurement. It can be said that the radioactive information of rocks has better or even diagnostic identification characteristics than the spectral information of rocks, especially in vegetation-covered areas.

Radioactive information of rocks can be obtained by airborne geophysical prospecting (airborne gamma spectrometry), which is similar to aerospace or aerial remote sensing, and is not limited by surface conditions, fast, efficient and economical. With the continuous improvement of the accuracy of airborne geophysical instruments, it is entirely possible to obtain more detailed information on the content of radioactive elements in surface rocks.

Of course, the radioactive information of rocks is by no means omnipotent in lithology identification. Just like "same spectrum heterogeneity, same spectrum heterogeneity" in remote sensing technology, the radioactive information of rocks also has the phenomenon of "the same radioactive element content but different lithology or the same lithology but different radioactive element content". Weathering and alteration of rocks can cause the loss, migration and enrichment of radioactive elements, thus leading to the change of radioactive characteristics of rocks. Therefore, it is not the best choice to use the single radioactive information of rocks for lithology identification or classification.

Remote sensing information is multi-information, including reflectivity information of different wave bands, and aviation radioactivity information is also multi-information, including γ total amount, U, Th, K content and their ratio information, both of which have the advantages of fast acquisition, low cost, large amount of information and large coverage area. Theoretically, it can be proved that the fusion of remote sensing and airborne radioactive information can not only fully express the spectral information of rocks, including their texture and structural characteristics, but also fully reflect the radioactive characteristics of rocks and enhance the resolution of lithology, which has high application value in geological survey and mineral evaluation. This is a comprehensive technology of light (multispectral and hyperspectral remote sensing) and energy (airborne radioactive gamma ray spectrum). Liu Dechang et al. (1993) integrated airborne radioactive information and remote sensing information for the first time in the research of "Multi-source information integration technology and application with airborne radiation measurement as the main information source", that is, a geoscience information processing system consisting of a computer system, a digital image processing system, a geographic information system and a computer-aided mapping system was used to develop a synthetic image with both spatial texture information and radioactive information in spectral images, thus achieving the following objectives. They preprocess aviation energy spectrum data, including interpolation, gridding and gray scale conversion; (2) Coordinate registration of pre-processed airborne energy spectrum data and remote sensing data; ③ The registered remote sensing and airborne energy spectrum data are combined into a series of new images in RGB. Therefore, they divided the uranium-producing rock mass in Lianshanguan, determined the uranium metallogenic structural model in Lianshanguan area, predicted the favorable metallogenic area, and showed its good application effect.

Feng et al. (1997) studied the application of multi-source geological information technology in uranium metallogenic condition analysis and prospect prediction. On the basis of preprocessing and geometric registration of the original data, TM 1, 5,7 is taken as intensity I, U, Th and K as hue H, and saturation S is constant 255, which is transformed from HIS space to RGB space. In this image, the boundaries of the main strata are clear, and the linear and annular structures are well reflected.

Table 8. 1 Contents and ratios of K, U and Th in different rock types

Note: The contents of K, U and Th in magmatic rocks and sedimentary rocks in the table are as follows (1985); The contents of uranium and thorium in metamorphic rocks are based on Chen et al. (1980).

Li Jianfeng et al. (1999) made multi-source statistical analysis on the U, Th, K and total data of four bands of MSS and four channels of airborne gamma spectrum, calculated their correlation matrix, and then took the first three principal components after K- L transformation for RGB synthesis and histogram equalization, so as to get one with rich colors and good lithology identification effect.

Zhu Minqiang (2002) converted grid data into image data by SURFER software in his doctoral thesis "Research on Comprehensive Evaluation Technology of Uranium Mine in Sandstone Basin Based on GIS", and then registered it with TM data by ERDAS image processing software. Then, the first and second principal components PC 1 and PC2 after K- L transformation of TM data are combined with aerial energy spectrum U image data for RGB, and the combined image provides a basis for uranium ore prediction.

The operation process of remote sensing and aviation radioactive information integration technology can be roughly summarized as the following five steps:

(1) preprocessing and geometric registration.

(2) Calculate the correlation coefficient matrix U, Th, K, Tc, U/Th, K/U between remote sensing data and aviation data in each band.

(3) Carry out K- L transformation, Maunsel transformation or inverse transformation on related variables.

(4) Using the principal component of K- L transform itself or other variables for false color synthesis and image enhancement, so as to obtain thematic images.

(5) Supervise the classification of thematic images and output thematic maps.

8. 1.2 Integrated image processing and analysis of remote sensing and airborne radioactive information in Xiangshan area

The remote sensing data in Xiangshan area are refined by polynomial geometry correction and SFIM fusion, and the airborne radioactive data and remote sensing data are aligned to the same UTM (Clarke, 1886) projection system. These two types of images have the geometric registration conditions for comprehensive processing.

On the Erdas (8.6) software platform, the images of SFIM 1, 2, 3, 4, 5, 7 and K, U, Th and Tc are superimposed and combined to calculate the correlation coefficient matrix (Table 8.2).

Table 8.2 Correlation coefficient matrix of SFIM 1, 2, 3, 4, 5, 7 and K, U, Th and Tc data in Xiangshan area

As can be seen from Table 8.2, there is a certain correlation between remote sensing data (SFIM 1, 2, 3, 4, 5, 7), and there is a good correlation between airborne data K, U, Th and Tc, but there is little correlation between remote sensing data and airborne data. Vegetation is developed in Xiangshan area, and remote sensing data contains less information about rock characteristics, while vegetation and terrain (texture) information are more, while aerial data mainly reflect lithological characteristics, which is the internal factor that the two kinds of information are not related to each other and the significance of integrating the two kinds of information.

In order to effectively integrate the two kinds of information, the author uses the K- L transformation method to decompose multivariate variables into several new irrelevant variables with decreasing information, that is, by compressing the data dimension, the original information loss is minimized, and the purpose of extracting subject information is achieved.

Table 8.3 Eigenvalues of SFIM 1, 2, 3, 4, 5 and 7 and K- L transformation of K, U, Th and Tc in Xiangshan area

See table 8.3 for the eigenvalues of K- L transformation. As can be seen from Table 8.3, ①KL 1, KL2 and KL3 contain 99.97% of the total information; ②KL 1 has the largest amount of information, mainly from Tc, KL8 from Th, KL9 from U, and KL 10 from K; ③KL2 is mainly reflected in SFIM5 information, but it is also reflected in other bands. ④KL6 has a strong load on SFIM-5 and SFIM-7 bands, but its contribution is opposite, so it can be used to extract abnormal alteration information of water-bearing minerals. Therefore, the RGB composite images of KL6, KL 1 and KL2 (Figure 8. 1) are more interpretable than single remote sensing or aerial data images.

In Figure 8. 1, Quaternary alluvial deposits are red, linear and branched; Red sandstone of Nanxiong Formation of Upper Cretaceous, blue, embossed and thick-brained; Porphyry lava of the Lower Cretaceous Ehuling Formation, yellow-green, with medium-deep cutting blocks; Granite porphyry or rhyolite dacite porphyry is similar in hue to porphyry lava, but its shadow is more complicated. Sinian metamorphic rocks, light gray-green, moderately shallow-cut and densely feathered; Caledonian or early Yanshanian granite, yellow-green, irregular claw ridge, well-developed dendritic water system.

In addition, figure 8. 1 also shows the image of the wire-loop structure well.

Compared with the simple remote sensing image (see Figure 4. 1), the most obvious improvement of this image is that there are obvious differences between the tone of basement metamorphic rocks and basement granite and volcanic-intrusive complex, which improves the visual effect of the image. Although the spatial resolution of remote sensing data reaches 15 m, the grid spacing of aerial survey data is 300 m, and the definition of most lithologic stratigraphic boundaries is still insufficient. Because the lithofacies mapping of 1∶25000 has been carried out in Xiangshan area, the spatial resolution of the synthetic image is far from meeting the requirements of finer lithologic identification and alteration information extraction. Only when the accuracy of airborne geophysical instruments is obviously improved, the application of this composite image, which can express both spectral information and energy spectrum information, in Xiangshan area has practical significance. Therefore, the author did not further process or classify this image.