GISdevelopment.net ---> AARS ---> ACRS 1997 ---> Poster Session 3 Gold Mineralization Information Processing On TM Images He Guojin, Hu Deyong (China Remote Sensing Satellite Ground Sttion, Beijing 100086) Abstract Geological structural and lithological characters, minieralization and alteration as well as biogeochemical anomalies are the three types of information features that have been normally considered in the procedure of minerogenetic prognosis using remotely sensed data. By now, the method for structure features extracting has been widely, successfully performed. It is of course more difficult to achieve mineralized (alternation) information along with bio-geochemical anomalies from satellite-based images. As a type of field information, satellite-based remote sensing information is the result of all the objects on the surface interacting with electromagnetic wave. Thus, all the objects on the surface are information carriers in which geological anomalies are involved. This results in a weak presentation of geological anomalies in TM images. This kind of weak information is just the target that we usually operate on by image processing. For the reasons above, we propose an ideal named "micro-different information processing." Its general views are as follows: Feature information reflecting the common characteristics of natural objects should be dispelled, and the variation information should be identified. For the variation information probably indicate geological anomalies. Thereby, according to these viewpoints, vegetation and soil are not again regarded as only disturbing coverage for mineralization information extraction. On the basis of the ideal of "micro-difference information processing," Some tests have been done to obtain gold mineralization anomalies from TM images in more than ten areas of China, where the coverage is much different. The results have shown good indications of gold mineralized information or vegetation anomalies. This information is very clear in processed TM images, and has been proved to be highly accurate by fieldwork. Among them, several gold deposits have been discovered. Introduction Satellite remote sensing has provided us an effective method for geological mapping and mineral surveying. Geological structures and lithological characters, mineralization and alteration as well as bio-geochemical anomalies are the three types of information features that have normally considered in the procedure of gold minerogenetic prognosis using remotely sensed data. By lithological interpretation and structural analysis, geologists can get some supplementary information for gold exploration from satellite images. On geological studies using remote sensing technique, researchers have recognized various kinds of liner and circular phenomenon existed in the satellite images. This caused an extensive study on the relationship about minerogenetic geological background and the statistical information of liner-circular structures, and results in people's high attention to the ore-controlling function of these structures. Statistic shows that 909 of 1000 endogenic metal deposits in China are closely related to liner-circular structures, while gold deposits reach to a ratio of 99.22%. In the East of Shandong province, 39 among the 40 gold deposits are linked to various kinds of circular bodies (Xue Chongshen 1993). The high statistical density of geological structures has positive relation to gold anomalies. It is therefore possible to obtain some useful information for gold mineral exploration with spatial analysis of liner-circular structures and their mathematical frequency statistics. As far as the mapping of mineralization (alteration) and bio-geochemical anomalies using satellite images is concerned, researchers have taken great effort to the studies since Rowan succeeded to map the gossan distribution using Landsat-MSS data in 1977, and have achieved much accomplishment. Some works have been successfully done to abstract the alteration information is different bare areas with TM5/TM7 and TM3/TM1 (rowan 1982, Abrams 1984, Chzves 1989 and Yuan Jihui 1990 etc.), While the combining way of bands' ratio and Principal Components Analysis has also been carried out to measure the alteration information in areas with dense vegetations coverage (Huntington 1988, zhao Yuanhong 1990, Chen Guanghuo 1990 and Ma Jianwen 1993 etc.). Then Lv Huiping made in inquiry into the methods of bio-visible - near infrared spectra (0.4-2.5 um) of TM data, the spectral reflectance of altered rock is far different from that of unaltered rock. The altered rock has a very high reflectance ratio in Band5 (1.55-1.75um) of TM, and an absorption band in band 7(2.08-2.35um). In addition, it is possible to bring about the ecological variation of the plants due to the high content (or loss) of some mineral elements in soil, and results in change of plant's spectral characteristics. All above is the theoretical basis that we are according to abstracting gold mineralization (alteration) information and bio-geochemical anomalies from satellite remotely sensed data. However, as a spectral information, it is the result of all the objects on the surface interacting with electromagnetic wave. Thus all the objects on the surface are information carriers in which geological anomalies are involved. Compared to the common information, the mineralization (alteration) is very weak in TM images. On the basis of the knowledge above and having performed trails on gold mineral exploration in more than ten areas of China, Prof. Hu Deyong raised an ideal named "Micro-difference information processing" with which gold mineralization information can be abstracted directly from TM images. Its application in remote sensing geology and image processing have conducted on obvious economical result and a long-term social result. The General Ideal Of "Micro-Different Information Processing" Many researchers become used to understanding and analyzing remote sensing images with the views of conventional geology. This results in having ignored the own physical characteristics and information mechanism of satellite-based images. Based on information theory, "Micro-difference information processing" looks into the geological information characteristics existed in TM data. Its general view are as follows: About geological remote sensing information. As the description above, all the objects on the surface are information carriers because of the reason that remote sensing information is the result of all the objects on the surface interacting with electromagnetic wave. By substance transmission and percolation, geological bodies and structures under soil or vegetation coverage will cause some changes of the surface, such as physical characteristics and chemical component, water contains as well as ecological variation of plants. These changes may lead to a weak presentation in TM images, and they probably indicate geological anomalies related to gold mineral exploration. Therefore this kind of weak information is just the target that we usually operate on by image processing. In other word, the key issue of gold mineralization information extracting from TM images is to dispel the common disturbing information and achieve the weak variation information. About vegetation coverage. Some remote sensing geologists look upon vegetation as only a coverage to geological bodies or geological anomalies, and they emphasize to apply satellite-data acquired in winter to reduce the influence of vegetation. Meanwhile they advise to get rid of vegetation coverage in image processing. Nevertheless high content (or loss) of some chemical elements in soil may lead to ecological variation as well as change of plants' spectral characteristics. So, in the areas of heavy vegetation coverage, we should pay more attention to bio-geochemical effect on TM images and select the TM temporal data for distinguished or the growing appearance of vegetation is most different, namely bio-geochemical effect is most obvious. About application preprocessing of remote sensing geology. Similar to information recognition in signal processing, the signal (information), as well mean in TM data, is relative to "noise," and it will change with the application purpose. Some part of the data is signal in one application, to another applications, it is probably only but noise (non-contribution information). For the information (such as alteration information) we need normally appears very weak, and the common information, like heavy noise disturbing in signal processing, is what we unexpected in TM images. Therefore, it is very significant to perform image preprocessing in geological application of remote sensing. Its function can be regarded as that of restraining noise and improving "signal to noise ratio" (S/N) in signal processing model. Some Applications Of "Micro-Difference Information Processing" And The Results In principal, the approaches of mineral exploration and minerogenetic prognosis can be divided into two ways, namely drawing analogy and seeking anomaly. The former way is to establish a model based on the geological condition, the ore-controlling factors and the mineral exploration symbols in the already known areas, and then apply it to uncertain areas. The latter considers that the procedure of mineral exploration is to seek anomalies different from geological background. Depending on information theory, "Micro-different information processing" tries to obtain geological anomalies in TM data. It thus is an approach of seeking anomalies, but in respect to types' judgment of the anomalies, it usually uses the method of drawing analogy. As trails, the ideal has been successively applied to the studies of gold mineral exploration in more than ten regions in China, where the weather condition, vegetation coverage and topography are much different. The results have shown good indications of gold deposits and gold mineralization anomalies. This information is very clear in processed TM images, and has been proved to be highly accurate by fieldwork. Followed are the typical examples. Some area in inner-Mongolia. Here is in arid, semi-arid climate condition, and covered by losses in great depth. The information of gold mineralization anomalies that is expressed in purple-red color in processed TM images has been totally verified to be gold deposits or gold mineralized belts. According to this processed information, and by having done ground truth investigation in details on that, the geologists of Changchun Gold Institute of deduced that there had the prospect of 49.6-ton gold future reserves in this area, among them 4-5-ton gold reserves has been demonstrated. Some area with heavy vegetation coverage in Yunnan province. By performing "micro-difference information processing," the ultra-basic rock which is related to gold mineralization has been accurately abstracted in TM image, and the mineralization anomalies has also been enhanced. This anomaly information does coincide with the known deposit and the geochemical anomalies. Some area in liaoning province, north-east of China. Seven target areas have been located on the basis of processed TM image, and 70-100-tons of gold future reserves have been presented. Conclusion and Discussion In summary, we may say that there is a bright prospect for gold mineral exploration utilizing satellite remote sensing technique. For TM data has its own photographic theoretical basis and information mechanism, it is important for us to understand the features forming of TM images in the procedure of image processing and its applications. "Micro-difference information processing" is an ideal that we are according to abstracting gold mineralization information directly from TM images. It helps us locate mineral exploration targets efficiently. Its theoretical basis is : (1) Spectral anomalies caused by the mineralization (alteration) of soil and rock (2) Change of vegetation spectral characteristics due to vegetation variation brought about by bio-geochemical effect. However, no way of image processing is able to improve absolute content of a certain information in TM image. Information theory announces the law that information content of discrete to time. On the one hand, information is not well distributed at different period. On the other hand, the time longer, the information content larger. In remote sensing application, it is impossible for us to reach the complete information measurement of a target by using infinite time, therefore how to obtain the largest information content is definite time is what we should pay more attention to. As to geological application using image acquired from the same satellite sensor, it is an efficient approach to improve the absolute information content by selecting reasonable temporal satellite data.