(Institute of Aerogeophysical Remote Sensing Center, Ministry of Geology and Mineral Resources, Beijing)
Abstract: A unique image restoration method for airborne data is introduced. The main technical keys of this method are as follows: (1) Put forward the principle and theoretical basis of airborne data image restoration; Establish the image restoration processing flow of airborne data; Formulate the method of reconstructing data grid files; The restoration effect and error of aerial data images are evaluated.
Key words: airborne data, atmospheric background, image processing, image restoration technology.
I. Introduction
Since the square column NaI crystal entered the airborne integrated aviation station in the early 1970s, the sensitivity and effectiveness of airborne radioactivity measurement have been significantly improved, and the demand of geological and geophysical exploration circles for airborne radioactivity measurement has undergone fundamental changes.
In the application practice of recent 20 years, it is difficult to correct the atmospheric radon background, which has always been the main difficulty that puzzles the application effect of this method. As a result, stripes appear on the drawings, which seriously affects the usability of the drawings and the effect of the method. The reason [1] can be summarized as follows: the radioactivity measured in the air not only comes from underground, but also is influenced by aircraft hardware environment, cosmic rays, radon and its daughters in the atmosphere. The latter is called atmospheric background interference, which is influenced by climate, wind, wind direction, temperature, season and when it is measured in a day. The main performance of atmospheric background interference is that the background level is different between sorties. The uranium channel is the most disturbed, followed by the potassium channel. Thorium channel and total channel are small, but they can't be ignored (see figure 7, figure 3 and figure 4 of color swatch). Because of this noise, information from geological bodies is often submerged in the noise. Fig. 3a (color version of fig. 7) is a three-element restored image of K (red), Th (green) and U (blue) in Hami Hill Survey Area, fig. 3b is a composite image of aviation raw data in this survey area, fig. 4a (color version of fig. 7) shows the early and late calibration readings of each flight, and fig. 4b is a total track raw data image. The existence of banded noise can be compared to colored curtains hanging in front of useful information images. The severity of banded noise makes it impossible to draw isoline map from the original aerial data of this work area.
The problem of air binding is a "worldwide" problem [2]. In Canada, the background correction is carried out by using the measurement results over scattered lakes, rivers and other waters, and good results are obtained without using the upward probe [1]. Aerogeophysical companies such as Geometrics Company of the United States rely on upward probe measurement as the basis of background correction [3]. Grasty[4] put forward in 1986 that when there is no lake in the survey area, the average value of the abnormal area on the survey line can be used instead of the background.
The method introduced in this paper is completely different from all the methods adopted in the world. This method can be called aerial image restoration technology in digital image processing. The main purpose of image restoration technology is to improve the given image. Restoration is a process of trying to use some prior knowledge of degradation to reconstruct or restore degraded images. Therefore, the restoration technology is to model the degradation and restore the original image to some extent by using the opposite process.
Dr. Cannon [5] studied an image restoration technology or pattern removal technology, which is suitable for extracting fingerprint patterns from regular patterns (such as textiles), improving defocused images, eliminating noise between satellite image detectors, and making images blurred by camera or object translation clear during exposure. Srinivasan also reported this kind of research [6]. Zhang Yujun et al. studied the image restoration of uneven illumination degradation in deep-sea manganese nodule photos [7]. The image restoration of aerial data is another example of the successful application of digital image restoration technology in geosciences, but the degradation of aerial data images is different from the above examples. After successful research, this method has been verified in six survey areas.
Second, the principle and theoretical basis of aerial data image restoration technology.
What is measured by aerial photography is a degraded image G(x, y), which can be regarded as the superposition of real image F(x, y) and interference image η(x, y) to simplify the degradation process, as shown in figure 1. The prior knowledge of aerial image degradation comes from the analysis of aerial measurement process and original image. In the process of measurement, the useful information from geological bodies does not shift with time. Interference is essentially time-varying, but it has become a function of (x, y) in the image because:
Zhang Yujun on new methods of geological exploration.
Figure 1 Schematic Diagram of Image Degradation of Aerial Ray Data
The change of η can be divided into jumps between sorties and gradual changes within sorties, as shown in Figure 4 (attached Figure 7 for color version). In each survey line, this interference is approximately constant. If x (that is, the column on the image) represents the direction perpendicular to the survey line, η(x, y) is simplified to η(x), and there are
Zhang Yujun on new methods of geological exploration.
The purpose of aerial image restoration is to approach η(x) as much as possible, so as to approach F(x, y). For this reason, the original images are listed many times along the measuring line.
Multi-line narrow and long window convolution;
Zhang Yujun on new methods of geological exploration.
Where w is a convolution template and a matrix composed of weighting factors. Convolution process is a linear operation, and its operator h does not change with space. Because the operator is linear, the response of the sum of two inputs is equal to the sum of two responses.
Zhang Yujun on new methods of geological exploration.
Since it is assumed that η is only related to x and the convolution window is single column, there are:
Zhang Yujun on new methods of geological exploration.
Now, analyze the properties of HF(x, y). Due to the multiple moving averages along the Y direction, local anomalies are "submerged" in the near-field characteristics, showing a low and gentle change along the survey line. If f(x, y) is used to represent local anomalies and L(x, y) is used to represent near-field fields, then:
Zhang Yujun on new methods of geological exploration.
Then the following processing is performed.
Zhang Yujun on new methods of geological exploration.
It can be seen from Equation (9) that the restored image f(x, y) is close to the real image from the local abnormal angle after the noise image is subtracted from the original image, and the error depends on the fluctuation range of the subtracted "near-area background value" in the survey line direction.
Thirdly, the process of airborne data image restoration.
The research of airborne data image restoration technology is based on multivariate statistics and takes image processing as a tool, which embodies the characteristics of fast and intuitive image processing, and its flow chart is shown in Figure 2.
Fig. 2 Processing Flow of Image Restoration of Aerial Photographic Data
In this method, it is assumed that the noise floor of the airborne amplifier is constant or changes linearly along the line direction. By moving the average along the survey line for many times, the local anomalies are gradually submerged in the noise background, and a noise image linearly related to the noise background is obtained. Noise image still needs edge effect compensation; The denoised image is finally restored by median filtering of spatial variables and contrast enhancement. This recovery process is shown in the left half of Figure 2.
The right half of Figure 2 is the reconstruction process of data grid files, which is essential for practical application. After classification and partition, the mean vector of each category before and after restoration is obtained, and the element content or counting rate of the restored image is obtained by least square fitting, and the grid file for drawing isoline map on the main computer is re-established.
In this study, the average value along the survey line is used as the noise level, and the result is not as ideal as the above method.
Fourth, the effect and error evaluation
1. Effect of image restoration of airborne data
Improvement of (1) drawing visual effect.
It can be said vividly that the restoration of aerial images is like opening a curtain with a cloth strip, so that the image that was faintly visible through this curtain reveals its true face, as shown in Figure 3a (attached Figure 7 for color version). The improvement of the visual effect of the drawing is also manifested in the elimination of the jagged noise on the rock boundary caused by the positioning problem, as shown in Figure 5 (Figure 7 of the color plate). Fig. 5 is a total trajectory contrast image, 5a is the original data, 5b is the noise image, 5c is the denoised image, and 5d is the restored image.
(2) The contour map made with the recovered data is true.
Taking Hami hill survey area as an example, due to the stripe interference of the original data, the isoline map of potassium, thorium and uranium channels cannot be drawn on the main computer, and only the plane profile map is provided; Only the main trajectory provides a contour map, but the influence of stripes can still be seen.
After image restoration and reconstruction of the grid file, it is fed back to the main computer to draw TC, K, Th and U isolines. Now, take the isolines of k recovered data in Figure 6 (color version of Figure 7) as an example. Compared with the geological map, the anomalies correspond well with geological bodies, and the radioactive trends of various lithology are consistent, which confirms the reliability of these isolines. The classification map made from the restored image also confirms this point, as shown in Figure 7 (attached figure 7 for color version). The numbers in Figure 7 are: ① ultrabasic rocks; ② basic rocks; ③ Granite; ④ diorite; ⑤ metamorphic rocks; ⑥ migmatite; ⑦ Quaternary sedimentation; (8) Tertiary and Quaternary sediments; Pet-name ruby tertiary deposits.
(3) Added useful information.
In this study, multivariate statistical method is used to quantitatively evaluate the restoration effect of aerial images. The image can be quantitatively evaluated by the magnitude of variance value composed of useful information. So we have to calculate the average of the total variation of the whole picture of a pixel, that is, the average variation value. With c, c? And c "respectively represent the average change value of useful information in the original image, the average change value of interference information in the original image and the average change value of useful information in the final restored image. Statistics in g? (x, y) approximately represents η (x); Use [G(x, y)-G? (x, y)] approximately represents f (x, y); The final restored image is represented by P(x, y), assuming no interference.
Zhang Yujun on new methods of geological exploration.
In the formula, the letter plus "-"indicates the average value; M and n are the number of rows and columns of the image.
Table 1 is the statistical result of quantitative evaluation of aerial data images in Hami mound investigation area according to the above categories.
Table 1
As can be seen from the table 1, the useful information of K, Th, U and TC has obviously increased after image restoration. As far as this work area is concerned, the relative quality of TC and K original images is good, while Th and U are poor.
2. Accuracy and error evaluation of restored images
The main error source of the restored image is the "near-area background value" L(x, y) formed by multiple moving averages. By counting the contour data of the interference image, the following accuracy evaluation is obtained:
K 0. 16%% (absolute content); th+2. 1 PP m；
u 0. 15 ppm； TC 869.6 counting.
Verb (abbreviation of verb) conclusion
(1) The method introduced in this paper is a unique airborne data and image restoration technology proposed for the first time at home and abroad, and its reliability and practicability have been verified in many work areas.
(2) This technology can basically eliminate the banding phenomenon caused by the change of atmospheric background and threshold, basically restore the true appearance of aerial images, and prepare for further image processing (such as derivation, enhancement, classification, logical operation, etc.). ), so this technology is also a fast pretreatment method.
(3) This method improves the image noise problem of jagged edges of some geological bodies caused by flight positioning displacement.
(4) In this study, the "average variance value of useful information" is established as a measure to quantitatively evaluate the image restoration effect of airborne data. As a pretreatment process, the possible absolute error of element content value or the accuracy of the method are also discussed.
refer to
[1] Grassy, R.L., Gamma-ray Spectrometric Determination in Uranium Exploration-Theory and Operating Procedures, Geophysics and Geochemistry in Metal Ore Exploration, GSC, Ottawa, 147- 162, 1977.
[2]screen, A.A. "Leveling airborne gamma radiation data with correlation information between channels", Geophysics, 52, 1557- 1562, 1987.
[3]Foote, R.S., "Improving the analysis of airborne gamma radiation data by eliminating environmental and soil radiation changes", in "Application of nuclear technology in mineral resources exploration and development: International Geoscience Conference, 2002". IAEA Mtg. Buenos Aires, 187- 196, 1968.
[4]Grasty, R.L., Automatic System for Calculating Online Atmospheric Background, GSC paper, 1-52, 1987.
[5]Cannon, M., Lehar, A. and Preston, F., Removing background patterns by power spectrum filtering, Applied Optics, 22, 777-779, 1983.
[6]Srinivasan, R., Software Image Restoration Technology, Digital Design,16,4,27-34, 1986.
Zhang Yujun, Shi Jianwen. Study on image restoration and image processing technology of deep-sea polymetallic nodule photos. Geophysical and geochemical exploration, 1989, (13): 435 ~ 44 1.
Thanks to Comrade Lin Zhenmin for his valuable comments on this article. Comrade Shi Jianwen took part in the repeated work area test, and comrades worked out the grid file conversion and least square fitting programs respectively. Comrade Yang took a screen picture, and Comrade Shui Enhai collected the calibration data of the test work area. Thank you.
Research on image restoration technology of airborne radiation data
Zhang
(Institute of Aerogeophysics and Remote Sensing Center, Ministry of Geology and Mineral Resources, Beijing)
abstract
In this paper, a specific method for restoring aerial radiation data images is proposed. The main technical keys involved in this study are: the advancement of principles and theories; Establishment of processing flow chart; Formulate the method of reconstructing gridded data files; Evaluation of recovery results and errors involved in recovery processing.
Keywords airborne radiation data, atmospheric background, image processing, image restoration technology.
Journal of Geophysics, 1990, Vol.33, No.4.