(A) Selection of Remote Sensing Information Sources
There are many remote sensing information sources available for the project, such as LandSat TM (ETM+), Spot data, Ikonos and so on. Considering the accuracy of mapping (1:1 million, 1:250,000, 1:100,000, 1:50,000), as well as the large scope of the work area and the short work cycle, TM (ETM+) data, which has a fast acquisition speed, medium resolution, a large number of bands, and an affordable price, was chosen as the basic information source. In view of the region's cold climate, long winter and short summer, long snow cover time and other characteristics, the time-phase as much as possible to choose in the information-rich, low cloud content in September to November, at the same time as far as possible to select the time-phase of the different scenes between the TM (ETM +) data is close to the same, and give full consideration to the seasonal changes caused by differences in geomorphology differences. For example, compared with the alpine area and the low warm area, the alpine area can be biased towards summer, and the low warm area can be biased towards spring and fall. This time, we obtained 95 views of TM (ETM+) data for the whole Tibet, and the time phase is mainly from 1999 to 2001.
Figure 2-1 Workflow Diagram
TM contains seven bands from visible to mid- and far-infrared spectral regions, of which the spatial resolution of the other six bands is 28.5m, except for TM6 spatial resolution of 120m in the far-(thermal-)infrared spectral region. 1:500,000 high-quality images can be obtained after image processing, and the amplification can ensure the accuracy of 1:250,000.ETM + (Landsat 7) panchromatic band spatial resolution of 15m, after fusion processing can get 1:50,000 high-definition images, fully meet the requirements of this remote sensing survey.
(2) TM image production
1. Image production process and spectral testing
TM image production is the basis of this remote sensing survey, and the main task is to provide high-quality remote sensing image data. The image production process is shown in Figure 2-2.
Figure 2-2 Remote Sensing Image Production Process Flow Chart
To study the remote sensing characteristics of different types of lakes and salt minerals, the project team carried out a wave spectrum test. The test minerals include rock salt, manganese, borax and chert, quanhua (carbonate rock), and the types of test lakes include freshwater, low-mineralization brine, and mineral-containing high-mineralization brine, etc. (Plate 1; Figure 2-3). The testing task was undertaken by the Petroleum Exploration and Development Research Institute of China National Petroleum Corporation (CNPC). The testing instrument was a near-infrared mineral analyzer with a minimum spectral resolution of 7 μm, a signal-to-noise ratio of 1/3,500 to 1/4,500, a spectral sampling spacing of 2 μm, and a testing speed of 40 s/time. From the thousands of values tested by this instrument, the results are stable and reliable.
Figure 2-3 Spectral curves of solid salt sediments from the salt lake in Tibet
The test results show that different types of lakes exhibit obvious differences in reflectance in different wavebands. According to the results of the spectral test of each water body samples: in the visible wavelength band, the reflectance change is not big, the reflectance curve is relatively straight, while in the near-infrared wavelength band (1.3 ~ 2.5 μm), its reflectance not only in the overall performance of the obvious differences, and reflectance curves also show different characteristics, but with the wavelength of each body of water increases, the overall decline in the trend of the performance of the same. Salt lake reflectivity is the highest, generally 50% to 80%; salt water lake, generally 30% to 50%; and freshwater lakes (or pure water) is the lowest, mostly less than 30%. At the same time, different mineralization of salt lake water samples testing shows that the reflectivity and mineralization is basically a positive relationship.
The reflectance curves of different types of lakes are also different. Salt lake curve is generally manifested as a small amplitude, its peaks and troughs of the fluctuation value of 3% to 5% between; saltwater lakes in about 5%; freshwater lakes up to about 10%. Saltwater lakes show small amplitude and low frequency, saltwater lakes show large amplitude and low frequency, freshwater lakes show large amplitude and high frequency.
2. Color Synthesis
In order to obtain the best synthetic remote sensing image, on the basis of wave spectrum test analysis, it is necessary to carry out wave band correlation analysis (Table 2-1) and variance statistics on the image. The results of the variance statistics of each band show that the TM3, TM4, TM5 and TM7 bands are richer in information, among which TM7 contains more information. the coverage of the brightness values of the TM1, TM2, TM3 and TM6 bands is narrower, and the contrast of the images is poorer. In addition, we found that the mean gray value of TM1 band is higher than that of TM2, TM3 and TM4 bands. The correlation analysis of the seven bands of TM shows that the correlation coefficients of TM1, TM2, and TM3 are greater than 0.894; the correlation coefficients of TM5 and TM7 are 0.985; the correlation coefficients of TM4 and each band are relatively small, with the maximum of 0.843; and the correlation coefficients of TM6 and each band are weaker, with the maximum of 0.742. That is to say, the stronger band combination is the one between TM1 and TM4. In other words, among the 7 bands of TM, the stronger band combinations are TM1, TM2, TM3 combination and TM5, TM7 combination, while TM4, TM6 bands have stronger independence and smaller correlation with other bands.
Table 2-1 List of correlations of TM bands
Through comprehensive analysis, considering that the TM6 band mainly reflects the emitted radiation information of the feature, and the main application in the study is the reflected radiation information of the feature, so TM6 is not involved in the color synthesis. According to the above analysis, in principle, one band should be selected from each of the three groups of TM1, TM2, TM3 combination, TM5, TM7 combination and TM4 for RGB synthesis. Two image map synthesis schemes, TM4, TM3, TM2 and TM7, TM4, TM1, were selected in the work. From the synthesis effect, both schemes reflect the main geological and mineral information such as lake and salt deposition more prominently.
3. Geometric fine correction
The purpose of geometric fine correction is to correct the image distortion caused by systematic and non-systematic factors in the process of image acquisition, so as to realize the geometric integration with the standard image or map. The specific steps are as follows:
First, select the ground control point (GCP).
Ground Control Points (GCPs) have obvious and clear locational identification marks on the image, such as intersections or bends of roads and rivers, building boundaries, and farmland boundaries. Select 20 to 30 evenly distributed control points within each map to ensure spatial alignment accuracy.
Second, select the correction equation.
The quadratic polynomial is selected as the correction model to establish the relationship equation between image coordinates and reference coordinates. Using the selected control point coordinates, the polynomial coefficients are found by least squares regression:
Remote sensing mineralization prediction of salt lake mineral resources in Tibet
Remote sensing mineralization prediction of salt lake mineral resources in Tibet
The equation: (X, Y) is the original coordinates of the image element; (x, y) is the map coordinates of the image element with the same name; n is the order of the polynomial.
Calculate the root mean square error of each ground control point. By calculating the root-mean-square error for each control point, ground control points with large errors can be checked, and the cumulative overall root-mean-square error can be obtained. An image size of 0.5 is specified as the maximum acceptable total RMS error, and if the actual total RMS error of a control point exceeds this value, the ground control point with the maximum RMS error needs to be deleted or updated, and if necessary, a new control point can be selected until the acceptable accuracy requirements are met.
Third, image element resampling.
The nearest neighbor method is chosen to resample the image.
Fourth, image alignment.
Geometric alignment is mainly for the different bands of data that have been fine correction, so that the position of the image point of the same name in different bands is accurately aligned, in order to avoid the position of different bands between the image element to avoid the impact of the accuracy of the production. There are two methods of image-image alignment and graphic-image alignment, and the method of graphic-image alignment is used this time.
4. Image Output
The computer digitally processed image is divided into 1:50,000, 1:100,000, 1:250,000, and 1:500,000 standard frames, and a color plotter is used to print out the image. The resolution of the plotter (HP3500CP, HP5000PS) used was 600DPI, 1200DPI, all with color reduction function, which can well reproduce the true color of the original RGB image, and ensure that the texture of the image is clear. After the image is output, the method of cold lamination and pressing film is used to attach a film to the paper image map, so that it can be waterproof, sunscreen and folding, which is conducive to the use of the field and indoor.
5. Image Quality Review
This work*** produced 95 remote sensing image maps with different accuracy. In general, the different combinations of images are rich in information, with delicate and coordinated image tones, moderate contrast, rich layers, prominent main information, sharp and eye-catching, small distortions, and obvious differences in the colors of different features, which can meet the requirements for mapping. However, due to the influence of unfavorable factors such as large area, complex terrain and diverse climate in Tibet, there are still some problems in the synthesis effect. Such as the eastern region of Tibet, especially the Yarlung Zangbo Grand Canyon area generally contains high cloudiness, while the forest vegetation development, the image effect is not good, the remote sensing of lakes, geology and minerals interpretation is not ideal.