Image processing of airborne
shortwave infrared split-spectral scanning data in detecting oil pools by
using Remote Sensing techniques Li Jiahong, Zhang
Jiangzhong, Yang Hong, Zhu Boqin Institute of Remote Sensing Application, CAS Abstract Hydrocarbon content anomalies in soil above oil pools are a good direct indicating index on the earth surface. It was confirmed that the index possessed two absorption peaks at the wavelength 2.31mm and 2.35mm by spectral measurement in the laboratory and in the field. Oil Remote Sensing anomalies were obtained by airborne shortwave infrared split- spectral scanning images through computer image processing based on spectral features of the index. These results was proved to be correspondent with the ascertained distribution of oil pools. Introduction N-12 oil field, located at the north-east of Xinjiang, had been chosen as the test-site (Fig. 1) to study the method of Remote Sensing image processing to oil exploration. The test site is in the front edge of an alluvial fan, which developed at the north foot of Tianshan Mountains, in east of Junggar Basin. It is adjacent to Guerbantonggute Desert and the terrain is almost even. In most part of the area there develop[ed saline-alkali soil and covered sparse vegetation; In a few marshlands there grew some bog plants. The test site is located at the north Bei-santai Palaeo-dome. A south-pitching fault named Beisantai fault spans the center of the area from the east to west. There developed two nose-like anticlines in the north of the fault. In this area it has been already exploited that the distribution and scale of the oil pool was coincided with two Permian nose anticlines and another U-like oil area in the south of Beisantai fault. From the aforementioned analysis this test site was considered as a typical one for the theme. Spectral Features of Hydrocarbon Materials and Analysis of ASISS Data 1. Analysis of Spectral Features of Hydrocarbon Materials Ingredient of hydrocarbon materials in soil was approximate to that of crude oil. The obvious absorption valley at the wavelength of 2.27-2.47 mm was found by spectral measurement. There were two peaks at the wavelength of 2.31mm and 2.35mm which were caused by the overtone and combination tone vibrations of –CH and –CH (fig. 2). The spectral features of crude oil and those soil samples above the oil pools in the laboratory and in the field were clearly stable and comparable. It has been proved that there were hydrocarbon content anomalies in soil in test site and the band at the wavelength of 2.27 to 2.46mm was very useful. The spectral characteristics of the index was the theoretical bases to oil exploration by applying Remote Sensing techniques. 2. Analysis of ASISS Data Airborne shortwave infrared split-spectral scanner is one of the best sensors to detect oil pools at present. Its work-band of 12 channels ranges from 1.5mm to 2.5mm. Eight channels were chosen in this experiment based on the analysis of spectral features. Parameters of the central wavelength, band width and detecting targets of each channel are described in Table 1. The table revealed that channel 6 and 7 were two spectral characteristic ones to explore hydrocarbon content anomalies in soil. Since the spectral overlap of the channels is high and so is the correlative coefficient among the channels data, the anomalies information as weakened. How to enhance, extract and distinguish the anomalies is one of the major problems to be solved in this paper. Image Preprocessing of ASISS data There excited radiant and geometrical deterioration and the distortion along scanning orientation in the test image whose size is 2-48 I * 256 c * 8 b. The image would not be correct unless it was preprocessed. 1. Image distortion along scanning direction and rectification Ground resolution of each pixel is different along the scanning direction on an image. the one of the pixel properly under the airplane is maximum : 30m * 30m. It decreased while scanning visual angel increased. According to the regular distribution pattern of resolution along the orientation, TANGENT transformation model was established and applied to finish the rectification of this distortion on test site images. Corrected image size was 2048| * 320 c 2. Radiative distortion and correction We picked image of band 6 as a example to discuss the distortion (image – 1). It showed that the left side was dark was the right was bright; in accordance with DN profile distributive cure of the identical earth surface features on the image, we discovered thath DN value was low on the left and high on the right. In the light of analysis of DN histogram distribution of the whole image, we also found that its distribution was not normal. From these three aspects, it was confirmed that the original band 6 image was clearly distorted by radiant, so were the others. The NORMAL method was used to redress the distortion. Its radical principle was summarized as follows : Firstly, respecting average and mean square deviation for each band was designed. Secondly, image distortion curve was acquired by analyzing statistics results of the whole image; Thirdly, DN value of adjustment was calculated in column order. At last, pixel DN value of the whole image was resampled with the adjustment. In the processing, we set the respecting means 128 and deviation 15 for all bands on the basis of balance analysis of statistics results for each band. We obtained radiant rectified images that wee comparable each other (Image -2 ). 3. Image geometrical deformation and rectification Flight posture change was an important one of the factors that made image distort. Usually GCP transformation was carried out to solve this problem tieh topographic map. There were few stable and obvious geography marks or targets in this area, meantime ASISS image ground resolution was extraordinary low (only 30M * 30M), so few ground control points (GCP) which related to image were determined, and we couldn’t rectify the image with topographic map. Aerial photomap which was selected as a image related to the ASISS image named slave image was geometrically corrected and mosaicked by optical processing. From the master and slave image, 25 GCP’s were selected and GCP file was established, and then image geometrical deformation was corrected by resembling with bi-linear interpolation method. Extraction of Anomalous information and Analysis of Anomalous Results. Shortwave infrared Remote Sensing data had a high correlativity nature because the high overlap of the scanner channels. The correlation coefficients between each two bands were almost over 0.80; Especially between band 3 and 4, the coefficient was high up to 0.98256; and the lowest one was 0.678998. The correlation weakened furtherly the anomalous information which was mainly included in band 6 and 7. So on the image by composing band 6, 7, 7, (R, G, B) the oil-pool Remote Sensing image anomalies were almost not displayed. FACTOR analysis of R type model can remove the correlation, make the anomalous information outstanding, and reach the aim of recogonation and extraction. Let X =[x1, x2, ……Xn]T as n-dimensional vectors that come from n bands, Y = [y1, y2, … …yn]T as n-dimensional vectors that come from n components and re decided by: Y = A*X The A is a load factor matrix of n*n, and come from orthogonal transformation of relative coefficient matrix R = {rij} n*n according to eh biggest variance principle. A = {aij] n*n, aij is load factor value of the band j to component, i, that is contribution of variance, meantime, it represents the amount of information of band j in component i. So the load factor matrix A become the important basis of selecting component variate and extracting the anomalous information. The image was transformed by FACTOR model and the images of 8 components were acquired individually. The load factor matrix us illustrated in Tab. 2. The selection of components and extraction of anomalous information was according to the spectral features of anomalous index and the numerical value of load factor. From Tab.2 the load factor of the original bands in components was analyzed. It was found thath band 6 and 7 contribute to components 3 and 4 greatly and to the others small. This illustrated that the component 3 and 4 included most of anomalous information that was in the orgional badn 6 and 7. Oil anomalous image (Image – 3) was gotten by selecting components 3 and 4 to make the colour composite 2,4,4 (R,G,B). The low absorption of hydrocarbon content nomalies in soil in characteristic bands showed dark black in the anomalous image. in the other words, eh dark color reflected the materials that behaved low absorption in the bands of 2.30 and 2.33mm. And from the spectral analysis of materials (vegetation, carbonate minerals, day and hydrocarbon) in test site only hydrocarbon materials have these spectral features. Comparing and integrated analyzing oil Remote Sensing anomalies image and the distribution of the known oil pool (Fig. 1), it was certified that the pattern of oil anomalies in the image was coincided with the distribution of the known oil pool. Conclusion Hydrocarbon content anomalies in soil are an important index for detecting oil resources by using Remote Sensing techniques. The low reflection by hydrocarbons at the wavelength of 2.31 and 2.35mm is the basis for extracting oil Remote Sensing anomalies. The shortwave infrared split-spectral scanner that included these two bands of 2.30, 2.33mm provided the technical promise. The image preprocessing and extraction and recognition of hydrocarbon content anomalies made direct detecting oil reservoirs become reality. References Zhu Zhenhai etc. Integrated eveluation fo Remote Sensing to oil-gas exploration, chinese science and technology press, 1991.
Table 2: The load factor matric by R-type FACTOR transformation
|