Deposit Image Model and optimization of Prospect Targets
Cui Zhenkui, Huang xianfang, Zhu Dai
Luo Fusheng, Gao jun, Wang Guojuan
(Beijing Research Institute of Uranium Geology, Bejing, China)
Jin Zhenguo, Yang Xiaoli, Fu Song
(Computer Application Institute of Nuclear Industry, Bejing, China)
Abstract
Based on the practice in the application of remote sensing techniques, a new concept-deposit image model has been put forward. It is a very important tool for regional prospective prognosis. Application of deposit image model has opened up a possibility for quick and effective selecting of promising areas.
This paper takes Beishen region for example to illustrate that remotesening information plays an important role in gold reconnaissance. The procedure are as follows : first, selecting resonal multispectral band combinations and image processing methods in the light of theory analyses and image processing tests with the purpose to get ideal enhancement, feature extraction results and recognition effectiveness ; then determining major ore-controlling factors (including structures, formations, dikes, alternations and son on ) and extracting their comprehensive image features, on the basis of geological interpreation in combination with regional investigation and detailed geological research on deposit- deposit dessection”; Finally summarzing recognition criteria, establishing deposit image model, and optimizing target areas.
The practice has shown that using deposit image model to optimize target areas is a scientific and successful method. Of the eleven predicted areas, three have been proven to be favourable where geochemical anomalies and gold mineralization have been found at the surface and good results have been obtained.
Introduction
How to apply remote sensing to mineral deposit prospecting is a subject which has been pursued for years. In the past, the work was done by comprehensive analysis (basically, geological analysis ) of some remote sensing interpretation information . Sure it is one of the methods to select promising areas. But how to uitilize the superiority of remote sensing (remote sensing signature are chracterized by objectiveness, directness, and vastness) to determine comprehensive
image features of various ore-controlling factors including structures, formations, dikes , and alternations, and establish deposit image model, then apply analogue method to select favourable areas is still a problem to be scrutinized recently and in the future.
Methods and Procedures
Research on deposit image model is a sythetical project which is based on understanding all ore –controlling factors and typical image signature interpretation. On one hand, one must pay attention to study on regional metallogenic conditions, “deposit dissection master gold ore-controlling factors to enhance and extract useful image information as much as possible. the procedures are described below (Fig. 1) :
Figure 1 Procedure Diagram of optimization of targets
1. Image Processing
Image processing is the basis for establishing deposit image model. Inorder to get typical image signatures, various image enhancement and quantiative image processing methods are employed in accordance with difference objects.
A. Image enhancement of rocks and alternations
(1) Colour composite
Colour composite plays an important role in discrimination various rock types, particularly gold source formation. Since image results depend on the selection of wave band, great attention has been paid to wave band selection. three factors affecting the image results are considered.
(A) Spectral characteristics of surface objects
Two factors have been taken into consideration : the first is division of rocktypes. TM band 7 and band 5 are selected. Since in the range of band 7 and band 5, diferent rock types have distinct reflectance charcteristics. The second is spectral charcteristics of altered rock. TM band 7 and band 4 are selected, because altered rocks in study area have an absorption band in band 7 and a valley in band 4
(B) Correlation coefficient
In order to make the selected bands represent all the original digital data, three bands which have the smallest correlation coefficient are selected according to the statistic results, the corelation coefficient between band 1 and band 7,4 issmaller, so band 1 must be added.
(C) Practical results of processed image
The goal of image processing is to get an ideal processed image in which different objects have conspicuous difference in colour hue, saturation and lamination.
(2) Principal components transformation
Principal components transformation is realized by calculating a new coordinate in order to get principal components which have smaller correlation coefficient. Usually the first principal component contains 85 percent of the original six bands of TM data, so the colour composite image made from the first three principal components does not simply represent the information of three bands, but contain almost all information of the initial six bands of TM data. subtal difference of surface objects, hydrothermal alternation and degrees migmatization can highlighted in the PC image. Some signatures which are not displayed in colour composit image can reflected in the PC image as well. So the method is called as “hidden information-highlighting method”.
(3) Ratio colour composite image
Ratio colour composite method plays a special role in identifying altered rock and accentuating “colour anomalies”. It is manipulated by exaggerating subtal difference of reflectance of different surface objects in individual band. On one band, ratio image can minimize difference in illumination condition, on the other hand, it can stretch the average value of different type surface objects with smaller intensity. In this way we can not only avoid taking the same type, but also improve the ability to distinguish different type objects with appoximate intensity.
B. Enhancement of linear strutures
Linear features can be enhanced by digital filters. Usually Laplacian transform is suitable for enhancing almost any orientation linear features. Spatial convolution is employed in emphasizing circular structures. directional filter is used to enchance specfic linear trends in the image, During the period od doing Laplacian transform, some experiences have been accumulated. According to the necessity, by changing kernel value, we can obtain a result image in which linear features have greatly been enchanced, while background image is obscured; and also prepare a result image in which original background image is maintained and linear features are enhanced.
