Research on breakthroughs in geological and mineral exploration and the application of technology for finding minerals

Band addition and subtraction combination operations can extend the difference in brightness value between bands, and the band ratio is the division operation between different bands of the image. The band ratio can highlight the features with different slopes shown by the spectral curves and can eliminate the multiplicative noise or factors between bands, which is one of the most commonly used and very effective methods for mineralized alteration information enhancement processing. As the result of alteration information extraction inevitably contains non-alteration remote sensing information, the “weak information” should be hidden in the background information and not extracted.

The ratio method is a multi-spectral image processing method, which uses the ratio of the gray scale of the corresponding image elements of two bands or several combined bands according to the spectral characteristics of the feature to extract the target information. The ratio method can expand the spectral differences of different features, eliminate or attenuate the interference information of terrain shadows, vegetation, etc., and enhance the extraction of weak mineralized alteration information.

Principal component analysis is often referred to mathematically as the KL transform, which is a multidimensional orthogonal linear transform in statistics. It is a multi-channel image in the same image field, the application of the KL matrix for the combination, so that the information in the original image is concentrated into the smallest possible number of new component images, so that these component images are not related to each other.

1)

Mathematical model of optimal density segmentation

For a single-channel image, assuming that the minimum value is Min(≥0) and the maximum value is Max(≤255) , all pixels are in the gray level range of {Min,Max} , and the number of pixels in each gray level can be counted and expressed as P_i.

If the segment is divided from the ith gray level to the jth gray level by {i,⋯,j}(Min≤i≤j≤Max) , the sum of the squared deviations within the segment (called the diameter of the segment) is: (8)D(i,j)=∑k=ij[k−M(i,j)]2×Pk

In the formula: (9)M(i,j)=(∑k=ijk×Pk)/∑k=ijPk

is the average gray value of the image elements within the segment, and ∑k=ijPk is the total number of image elements within the segment.

Using the above equation, the intra-segment sum-of-squares matrix of gray levels can be calculated D: (10)D=[ D( Min, Min ) D( Min, Min +1) ⋯ D( Min , Max ) D(Min+1,Min+1) ⋯ D( Min +1,Max) ⋱ D(Max,Max)]

Clearly, matrix D is symmetric and D(i, i) = 0, so only the upper triangles need to be computed.

There is a partitioning method which splits the Max − Min + 1 gray levels into k segments as follows: {Min,⋯,i2}{i2+1,⋯,i3}⋯{ik+1⋯,Max} . For each segment, determine the diameter of the segment separately as follows: D(Min,i2),D(i2+1,i3),⋯,D(ik+1,Max) . For this partitioning, the sum of the diameters of the segments can be S: (11)S=D(Min,i2)+D(i2+1,i3)+⋯+D(ik+1,Max)

The optimal density segmentation is to find the segmentation method that minimizes S among all possible k-segment segmentations, which is called the optimal k-segment segmentation. 2)

Optimal two-segment segmentation

The Max − Min + 1 gray levels are divided into two segments, there are Max − Min kinds of segmentation, which one is optimal, only need to calculate the sum of the diameters of various segmentation methods, and find the smallest sum of the diameters is the optimal two-segment segmentation. That is, for any i(Min<i<Max) can determine a two-segment segmentation method, {Min,⋯,i}{i+1,⋯,Max} , the corresponding two segments of the sum of the diameter is: (12)Si(2)(Max)=D(Min,i)+D(i+1,Max)

Where (Max) denotes the maximum gray level to be segmented. (2) denotes the number of pre-segmented segments. (2) denotes the number of pre-segmented segments. i denotes the segmentation with gray level i as the segmentation point.

It is already known that Si(2)(Max) is the within-group sum of squared deviations S_Inner of segment {Min, ⋯, i} and segment {i + 1, ⋯, Max}, and if the corresponding total sum of squared deviations and between-group sum of squared deviations are denoted by S_Total and S_Inter, respectively, we have, according to the principle of analysis of variance (ANOVA): (13)STotal=SInter+SInner

When the sum is minimum, the sum of the squared deviations between the groups must be maximized. Therefore, it is sufficient to find the appropriate i such that Si(2)(Max) is minimized, which is the optimal partition sought. Assume that Si(2)(Max) is minimized when i = i₂, i.e: (14)Si(2)(Max)=MIN[Si(2)(Max)]

Then {Min,⋯⋯,i2} is the first segment of the optimal partition; {i2+1,⋯⋯,Max} is the second segment of the optimal partition. The sum of squared deviations within the corresponding segments is: (15)SInner(2)=Si2(2)(Max) 3)

Optimal three-segment segmentation

For any j(Min+1≤j≤Max) , make an optimal two-segment segmentation for the first j − Min + 1 gray levels, notation: (16)Si(2)(j)=D(Min,i)+D(i+1,j)

For i find the minimum value as: (17)Si2(j)(2)(j)=MIN[Si(2)(j)] (Min≤i≤j−1)

Thus the optimal bisection for the first j − Min + 1 gray level is: (18){Min,⋯,i2(j)},{i2(j)+1,⋯,j}

Si2(i)(2)(j) denotes the sum of squared intra-group deviations of the corresponding two segments.

The optimal two-segment segmentation of the first j − Min + 1 gray levels (Eq. above) and {j+1,⋯,Max} form a three-segment segmentation to find a proper j such that: (19)Sj(3)(Max)=Si2(j)(2)(j)+D(j+1,Max) (Min+1≤j≤Max−1)

As small as possible, find the minimum value of Sj(3)(Max) for all j. Let Sj(3)(Max) be minimized when j = i₃, that is: (20)Si3(3)(Max)=MIN[Sj(3)(Max)] (Min+1≤j≤Max−1)

This determines an optimal three-segment partition: (21){Min,⋯,i2(i3)},{i2(i3)+1,⋯,i3},{i3+1,⋯,Max}

The sum of the squares of their corresponding intra-group deviations is: (22)SInner(3)=Si3(3)(Max)

Repeating the above steps in an analogous fashion leads to an optimal k-segment split: (23){Min,⋯,i2}{i2+1,⋯,i3}⋯{ik+1,⋯,Max} 4)

Determination of the optimum number of segments

The number of segments is determined by the curvilinear relationship between the sum of diameters S_ik(k)(Max) and the number of segments k.

Among them: (25)Si(k)(j)=Si(k−1)(i)(k−1)(i)+D(i+1,j) Min+k−1≤j≤Max;2≤k≤Max−Min+1

The optimal k-segment segmentation corresponding to any first j − Min + 1 gray level is found out step by step Sik(j)(k)(j) with the segmentation point ik(j), and from this, the optimal segmentation points for Max − Min + 1 gray levels of k segments (2≤k≤Max−Min+1) can be derived, and they are arranged as follows in descending order: ik(Max),ik−1(ik),ik−2(ik−1),⋯,i2(i3) . The corresponding diameters are summed up to be Sik(Max)(k)(Max) . The optimal k-segment segmentation is then: (26){Min,⋯,i2}{i2+1,⋯,i3}⋯{ik+1,⋯,Max}

Various types of etching information is always with some of the original band, the ratio variable or the main component factor has an “affinity” relationship, due to the etching information in the image of the spatial distribution of discrete, inevitably make two are related to a certain type of etching band image there is a spatial correlation. Based on the spatial correlation of such a premise can be the spatial distribution of etching information location and shape “carved” out.

The steps of the algorithm are as follows: 1)

Based on multivariate data analysis, select images related to certain types of etching.

Or: (28)Cosθij(h,l)=|∑k((xik−mi)⋅(xjk−mj))/∑k((xik−mi)2⋅(xjk−mj)2)| 3)

Transform Cos θ_ij(h, l) from 0 to 1.0 to 0 to 255.

The etching information is enhanced by the ratio transformation or KL transformation of the image, and the etching information is revealed from low degree of ordering to high degree of ordering. According to the principle of information entropy, the entropy of etching information must change in quantity. Therefore, with the idea of wavelet transform, the degree of orderliness is revealed by the method of calculating the information entropy of the window, and the objective existence degree of the etching information is estimated by the relativity of the degree of orderliness.

The degree of orderliness (R(t)) is defined as: (29)R(t)=1−H(t)/Hmax

where H(t)=−∑Pi log2Pi , Hmax=−∑(1/N)log2(1/N)=log2N .

The steps of the algorithm are as follows: 1)

Select a certain class of etching information to enhance the image (ratio or principal factor).

As the application of this technology to carry out the mineral search work needs to prioritize the extraction of remote sensing anomaly information, and then through the comparison of the extracted data with the data source, so as to clarify the specific types of underground minerals. Currently, the data sources used in geological and mineral exploration mainly include the American Landsat series data MSS, TM, ETM and the AS-TER data cooperated by NASA and METI of Japan. Both kinds of data have set up corresponding channels near the alteration mineral characteristic spectral band of 2.5 μm. By comparing the ASTER data with the ETM data, the data characteristic parameters of ETM and ASTER are shown in Table 1. It can be found that the band resolution of ASTER data is obviously higher, and the spatial resolution of band 1 and band 3 is 32. Therefore, the application of ASTER data for the comparison of remotely sensed alteration anomalies can obtain a higher identification accuracy. However, ETM data also has obvious advantages, and the coverage area of this database is obviously higher than that of ASTER data. The reconstruction of the ratio image can be realized by dividing the brightness values of the same image element in different bands. For example, the alteration minerals containing hydroxyl and carbonate have strong absorption in the 2.5μm band, so they will show low brightness on TM7, so in the actual work process, the relevant staff will apply TM5 or TM7 to effectively extract the information of the data containing hydroxyl or carbonate alteration minerals.

1)

Pre-processing

Useful information such as lithology and alteration minerals are weak information in the total information of the image, and are submerged in a large amount of conventional information such as vegetation, desert, host rocks, and so on. Therefore, it is necessary to eliminate interference in order to highlight the relevant geological and mineral information. The extraction of alteration information is severely impacted by the absorption of vegetation in the short-wave infrared interval. The extraction results are depicted in Fig. 3. Mask image production is adopted to try to exclude the interference of vegetation and desert on the weak information. The vegetation information is extracted by using the normalized vegetation index NDVI: (4-3)/(4+3). The histogram analysis of the produced NDVI image was performed to test and select the threshold value. Finally, a 0/1 binary image of the vegetation is produced based on the threshold value, which is used as a mask image to exclude vegetation interference. This image is multiplied with the information extraction result image to exclude false alteration information caused by vegetation. Using the same method, a 0/1 binary image of the sandy area was produced by taking advantage of the fact that the Gobi sandy area is in channel 13 and the reflectivity is higher than that of other channels.

Figure 3.

Vegetation and sand histogram and its threshold determination

Figure 4.

Base index anomaly information

Remote sensing information is used to find minerals in the Huhe Huduge area, and the analysis results are shown in Fig. 5, which synthesizes the remote sensing interpretation information and geological, physical exploration and chemical exploration data. The quartz-rich anomalies are developed in the north-northeast direction, located in the Agulu Gou Formation of the Jialtai Mountain Group of the Middle Paleozoic, controlled by two north-east-trending faults, and the northern part of the alteration area is adjacent to the Late Paleoproterozoic Permian diorite intrusion rock body. To the west, sericite, albite and illite anomalies are developed, located in the Agulugou Formation of the Middle Paleozoic Jialtai Mountain Group, controlled by north-east-northwest faults, and the west side is adjacent to the Late Paleozoic Permian-like mottled orthogneiss intrusive rock body. The southwesterly developing chlorite, green cordite, carbonate anomalies, divalent iron anomalies, quartz-rich anomalies are located in the Agulugou Formation of the Jialtai Mountain Group of the Middle Paleocene, controlled by north-east tectonic structure, and the east side is adjacent to the Late Paleozoic Permian speckled orthoclase intrusive rocks, which has a very good prospect of prospecting for minerals geologically.

Figure 5.

Call and huguger area for the mining analysis

From the perspective of spatial location, the alteration anomalies are mainly distributed in the Agulugou Formation of the Jhaltai Mountain Group, which is the main ore-bearing stratum, and the alteration anomalies are consistent with the regional stratigraphy and major tectonics, and are obviously controlled by stratigraphic tectonics, which indicates that there is strong hydrothermal effect in the area, and the anomalies are located close to or inside the magmatic rocks, which is the response to the alteration effect and the correlation of magmatic activities.