A method for enhancing remote sensing images in geological surveying and mapping

Abstract

The present invention relates to the field of image enhancement, and specifically to a method for enhancing geological surveying and remote sensing images. The method first obtains a multispectral image of a geological surveying and mapping area to be measured, performs color space conversion on the multispectral image, and obtains the hue channel value and brightness channel value of each pixel in the multispectral image; the multispectral image is segmented to obtain different geological areas and the category of each geological area in the multispectral image, and the geological area is segmented into superpixels, and the edge shape features of the superpixel block, as well as the distribution of the hue channel value and the change of the brightness channel value inside the superpixel block, obtain the non-natural feature coefficient of the superpixel block, and then denoise the multispectral image by obtaining the interference index of the ground object classification of each geological area to obtain an enhanced image. The present invention can achieve effective denoising of remote sensing images and improve the enhancement effect of remote sensing images.

Description

Technical Field

[0001] The invention relates to the field of image enhancement, and in particular to a method for enhancing geological surveying and mapping remote sensing images. Background Art

[0002] Geological survey remote sensing images refer to image data obtained from the air using remote sensing technology and used for geological research and analysis. During the collection process of geological survey remote sensing images, they are usually interfered by noise such as sensor noise or atmospheric noise, which causes the geological information in the geological survey remote sensing images to be blurred, which is not conducive to subsequent geological analysis. Therefore, it is necessary to further enhance the collected geological mapping remote sensing images.

[0003] In the related art, the collected geological survey remote sensing images are usually enhanced by using a filtering denoising algorithm with preset filtering parameters. However, the noise contained in the remote sensing images under different acquisition conditions is difficult to predict in advance, and the geological survey remote sensing images usually contain a large number of different types of geological areas. There are differences in the image characteristics of different areas after being affected by noise. As a result, the filtering denoising algorithm with preset filtering parameters cannot achieve effective denoising of the remote sensing images, thereby reducing the enhancement effect of the remote sensing images. Summary of the invention

[0004] In order to solve the technical problem that the filtering denoising algorithm using preset filtering parameters cannot effectively denoise the remote sensing image, thereby reducing the enhancement effect of the remote sensing image, the purpose of the present invention is to provide a geological surveying and mapping remote sensing image enhancement method, and the technical scheme adopted is as follows:

[0005] The present invention proposes a method for enhancing geological surveying and remote sensing images, the method comprising:

[0006] Acquire a multispectral image of the geological survey area to be measured, perform color space conversion on the multispectral image, and obtain a hue channel value and a brightness channel value of each pixel in the multispectral image; segment the multispectral image to obtain different geological areas in the multispectral image and a category of each geological area, wherein the category includes a natural category and a non-natural category;

[0007] Taking any geological region as a target geological region, performing superpixel segmentation on the target geological region to obtain multiple superpixel blocks of the target geological region, taking any superpixel block in the target geological region as a target superpixel block, and obtaining the shape regularity of the target superpixel block according to the direction of the pixel points on the edge of the target superpixel block; obtaining the color feature value of the target superpixel block according to the distribution of the hue channel value of the pixel points in the target superpixel block and the change of the brightness channel value of the pixel points; obtaining the non-natural feature coefficient of the target superpixel block based on the shape regularity and the color feature value;

[0008] Taking any geological region adjacent to the target geological region as a reference geological region, obtaining an initial interference index of the target geological region with respect to the reference geological region according to the position distribution of the centroids between the target geological region and the reference geological region, the non-natural feature coefficients of the superpixel blocks in the target geological region, the distance between the centroids of the superpixel blocks and the reference geological region, and the category of the target geological region; obtaining a ground feature classification interference index of the target geological region according to the initial interference index of the target geological region with respect to all adjacent geological regions;

[0009] Based on the ground object classification interference index of each geological area, the multispectral image is denoised to obtain an enhanced image.

[0010] Furthermore, obtaining the shape regularity of the target superpixel block includes:

[0011] Performing curve fitting on the pixel points on the edge of the target super-pixel block to obtain the edge fitting curve of the target super-pixel block;

[0012] Acquire a plurality of inflection points on the edge fitting curve, wherein any two adjacent inflection points divide the edge fitting curve into a curve subsegment;

[0013] Taking any curve subsegment of the edge fitting curve as the target curve subsegment, analyzing the overall level of curvature of all pixel points on the target curve subsegment and performing negative correlation mapping to obtain the first shape regularization parameter of the target curve subsegment;

[0014] Analyze the overall level of the slopes of all pixel points on the target curve subsegment to obtain the directional characteristic value of the target curve subsegment; use the absolute value of the difference between the directional characteristic value of the target curve subsegment and each other curve subsegment except the target curve subsegment as the directional difference between the target curve subsegment and each other curve subsegment; use the average value of the directional difference between the target curve subsegment and all other curve subsegments as the second shape regularization parameter of the target curve subsegment;

[0015] The first shape regularization parameter and the second shape regularization parameter are integrated to obtain an integrated shape regularization parameter of the target curve subsegment;

[0016] The average value of the comprehensive shape regularity parameters of all curve subsegments of the edge fitting curve is taken as the shape regularity of the target superpixel block.

[0017] Furthermore, obtaining the color feature value of the target superpixel block includes:

[0018] Analyze the discrete degree of the hue channel values ​​of all pixels in the target superpixel block to obtain the hue distribution parameter of the target superpixel block;

[0019] On the edge fitting curve of the target superpixel block, the curve subsegment with the largest length is used as the reference curve subsegment of the target superpixel block, and the directional feature value of the reference curve subsegment is mapped by using an inverse tangent function to obtain the texture direction angle of the target superpixel block;

[0020] The absolute value of the difference between the brightness channel value of each pixel in the target super-pixel block and the next adjacent pixel at the texture direction angle is used as the brightness change value of each pixel in the target super-pixel block; the average value of the brightness change values ​​of all pixels in the target super-pixel block is negatively correlated to obtain the brightness change parameter of the target super-pixel block;

[0021] The hue distribution parameter and the brightness change parameter are integrated to obtain the color feature value of the target super pixel block.

[0022] Furthermore, obtaining the unnatural characteristic coefficient of the target superpixel block based on the shape regularity and the color characteristic value includes:

[0023] The product value of the shape regularity and the color feature value is normalized to obtain the unnatural feature coefficient of the target super pixel block.

[0024] Furthermore, obtaining the initial interference index of the target geological region with respect to the reference geological region includes:

[0025] Connecting the centroid of the target geological region and the centroid of the reference geological region with a straight line to obtain a centroid line between the target geological region and the reference geological region, and taking a straight line passing through the centroid of the target geological region and perpendicular to the centroid line as a segmentation line of the target geological region, the segmentation line divides the target geological region into two sub-regions, and taking the sub-region adjacent to the reference geological region as a target sub-region of the target geological region;

[0026] Taking the ordered sequence composed of the non-natural feature coefficients of all superpixel blocks in the target sub-region as a reference sequence, and obtaining the segmentation threshold of the target geological region according to the difference between two adjacent non-natural feature coefficients in the reference sequence;

[0027] According to the non-natural feature coefficient of each super pixel block in the target geological region, the category of the target geological region and the segmentation threshold, super pixel blocks of different categories in the target geological region are selected from all super pixel blocks in the target geological region;

[0028] The absolute value of the difference between the non-natural feature coefficient of each non-same superpixel block and the segmentation threshold of the target geological area is used as the first abnormal coefficient of each non-same superpixel block; the Euclidean distance between the centroid of each non-same superpixel block and the centroid of the reference geological area is negatively correlated to obtain the second abnormal coefficient of each non-same superpixel block; the first abnormal coefficient and the second abnormal coefficient are combined to obtain the comprehensive abnormal coefficient of each non-same superpixel block;

[0029] The accumulated value of the comprehensive anomaly coefficients of all superpixel blocks of different categories in the target geological region is used as the initial interference index of the target geological region with respect to the reference geological region.

[0030] Furthermore, obtaining the segmentation threshold of the target geological area according to the difference between two adjacent non-natural feature coefficients in the reference sequence includes:

[0031] Any two adjacent non-natural characteristic coefficients in the reference sequence are regarded as an element group, and the absolute value of the difference between the two non-natural characteristic coefficients in each element group is regarded as the element variation of each element group;

[0032] The average value of two non-natural characteristic coefficients in the element group corresponding to the maximum value of the element variation is used as the segmentation threshold of the target geological area.

[0033] Furthermore, selecting superpixel blocks of different categories in the target geological area from all superpixel blocks in the target geological area includes:

[0034] If the category of the target geological area is a natural category, then in the target geological area, a superpixel block corresponding to a non-natural feature coefficient greater than a segmentation threshold is selected as a superpixel block of a non-same category in the target geological area;

[0035] If the category of the target geological region is a non-natural category, then in the target geological region, a superpixel block corresponding to a non-natural feature coefficient that is less than a segmentation threshold is selected as a non-same category superpixel block of the target geological region.

[0036] Furthermore, obtaining the ground object classification interference index of the target geological area includes:

[0037] The average value of the initial interference indexes of the target geological region with respect to all adjacent geological regions is used as the ground feature classification interference index of the target geological region.

[0038] Furthermore, the denoising of the multispectral image based on the ground object classification interference index of each geological region to obtain an enhanced image includes:

[0039] The ground object classification interference index of each geological region is used as a smoothing parameter used by a filtering denoising algorithm, and the filtering denoising algorithm is used to denoise each geological region in the multispectral image to obtain an enhanced image.

[0040] Furthermore, the obtaining of different geological regions and categories of each geological region in the multispectral image includes:

[0041] Acquire a plurality of multispectral image samples from a database, wherein the multispectral image samples contain regions with different annotated categories, and the annotated categories are divided into natural categories and non-natural categories;

[0042] The support vector machine is trained using multispectral image samples, and the trained support vector machine is used to segment the multispectral image to obtain different geological areas in the multispectral image and the category of each geological area.

[0043] The present invention has the following beneficial effects:

[0044] The present invention takes into account that the existing filtering denoising algorithm using preset filtering parameters cannot achieve effective denoising of remote sensing images, thereby reducing the enhancement effect of remote sensing images. Therefore, the present invention first collects remote sensing images of the geological area to be measured to obtain multispectral images, and performs color space conversion on the multispectral images to obtain hue channel values ​​and brightness channel values ​​of each pixel in the multispectral images. Since the multispectral images usually contain a large number of areas with different geological characteristics, the multispectral images can be segmented and multiple geological areas and the categories of each geological area can be extracted. Considering that the influence of noise will cause deviations in the above-mentioned segmentation results, that is, there may be areas of other geological types within a certain segmented geological area, first obtain multiple superpixel blocks of the target geological area, and then analyze the non-natural features of each superpixel block, and extract superpixel blocks with different categories from the target geological area from the target geological area, so as to accurately analyze the noise interference on the target geological area and improve the maximum The final image enhancement effect is obtained by considering that the shape of the superpixel blocks in the non-natural geological area is more regular than that of the superpixel blocks belonging to the natural geological area. Therefore, the regularity of the shape of the edge of the target superpixel block can be reflected by the regularity of the shape. At the same time, the color distribution consistency of the superpixel blocks in the non-natural geological area is poor, and the degree of brightness change is small. Therefore, the color feature value can be used to reflect the hue and brightness characteristics of the pixels inside the target superpixel block, and then the non-natural feature coefficient can be used to reflect the possibility that the target superpixel block belongs to the non-natural geological area. Considering that due to the influence of noise, the target geological area usually contains part of the adjacent geological area. Therefore, the initial interference coefficient can be used to further reflect the possibility that the part of the target geological area containing the adjacent geological area is caused by noise interference. Then, the ground object classification interference index can be used to reflect the degree of noise interference on the target geological area. Based on the ground object classification interference index, the multispectral image is effectively denoised to improve the image enhancement effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0046] Figure 1 A flow chart of a geological surveying and mapping remote sensing image enhancement method provided by one embodiment of the present invention;

[0047] Figure 2A flow chart of a method for obtaining an initial interference index of a target geological region with respect to a reference geological region provided by one embodiment of the present invention;

[0048] Figure 3 A schematic diagram of a target sub-region of a target geological region provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0049] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of a geological mapping remote sensing image enhancement method proposed by the present invention, its specific implementation method, structure, features and effects, in conjunction with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0050] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0051] The specific scheme of the geological surveying and mapping remote sensing image enhancement method provided by the present invention is described in detail below with reference to the accompanying drawings.

[0052] See also Figure 1 , which shows a flow chart of a geological surveying and remote sensing image enhancement method provided by an embodiment of the present invention, the method comprising:

[0053] Step S1: Acquire a multispectral image of the geological survey area to be measured, perform color space conversion on the multispectral image, and obtain the hue channel value and brightness channel value of each pixel in the multispectral image; segment the multispectral image to obtain different geological areas in the multispectral image and the category of each geological area, including natural categories and non-natural categories.

[0054] Since multispectral images can capture multiple spectral channels of objects at the same time, thereby providing richer and more detailed information about objects, multispectral images are usually used as geological mapping remote sensing images in the field of geological mapping and remote sensing. Therefore, the embodiment of the present invention first uses an unmanned aerial vehicle device equipped with a multispectral imager to collect a multispectral image of the geological mapping area to be measured, and then performs color space conversion on the multispectral image to obtain the hue channel value and brightness channel value of each pixel in the multispectral image. The hue channel value and the brightness channel value can be used to analyze the color and brightness characteristics of different areas in the multispectral image, thereby improving the final image enhancement effect. The color space selected for the conversion needs to include a hue channel and a brightness channel. For example, HSV color space or HSI color space can be selected. The color space conversion of multispectral images is a technical field well known to technicians in this field and will not be elaborated here.

[0055] Since multispectral images usually contain a large number of different geological features, such as vegetation, bare land, rocks, and farmland, etc., the embodiment of the present invention also needs to segment the multispectral image to obtain different geological regions in the multispectral image and the category of each geological region, wherein the categories of geological regions are divided into two categories, namely natural categories and non-natural categories. For geological regions of the natural category, it means that the geological region is a naturally formed region, such as mountains or vegetation regions, and for geological regions of the non-natural category, it means that the geological region is an artificially formed region, such as farmland or buildings.

[0056] Preferably, in one embodiment of the present invention, the method for acquiring different geological regions and the category of each geological region in the multispectral image specifically includes:

[0057] First, a plurality of multispectral image samples are obtained from a database, wherein the multispectral image samples contain regions with different labeled categories, and the labeled categories are divided into natural categories and non-natural categories; then, a support vector machine is trained using the multispectral image samples, and the trained support vector machine is used to segment the multispectral image, so as to obtain different geological regions in the multispectral image and the category of each geological region. Support vector machine is a technical means well known to those skilled in the art and will not be elaborated here.

[0058] Step S2: taking any geological area as the target geological area, performing superpixel segmentation on the target geological area to obtain multiple superpixel blocks of the target geological area, taking any superpixel block in the target geological area as the target superpixel block, and obtaining the shape regularity of the target superpixel block according to the direction of the pixels on the edge of the target superpixel block; obtaining the color feature value of the target superpixel block according to the distribution of the hue channel values ​​of the pixels in the target superpixel block and the change of the brightness channel values ​​of the pixels; and obtaining the non-natural feature coefficient of the target superpixel block based on the shape regularity and the color feature value.

[0059] Due to the interference of noise on the multispectral image, the above-mentioned image segmentation results will be biased. That is to say, within a certain segmented geological area, there may be parts of other geological types of areas. Usually, there are parts of non-natural categories in natural categories of geological areas, or there are parts of natural categories in non-natural categories of geological areas. Therefore, the embodiment of the present invention first analyzes any geological area, takes any geological area as the target geological area, and performs superpixel segmentation on the target geological area to achieve refined segmentation of the target geological area, thereby obtaining multiple superpixel blocks of the target geological area. Subsequently, the non-natural features of each superpixel block can be analyzed, and superpixel blocks of a different category from the target geological area can be extracted from the target geological area, thereby accurately analyzing the noise interference on the target geological area and improving the final image enhancement effect. Among them, superpixel segmentation is a technical means well known to those skilled in the art and will not be elaborated here.

[0060] Considering that the edge shapes of superpixel blocks belonging to non-natural geological areas, such as superpixel blocks in farmland or buildings, are relatively regular compared to those belonging to natural geological areas, any superpixel block in the target geological area is first taken as the target superpixel block, and the direction of the pixel points on the edge of the target superpixel block is analyzed. The regularity of the shape is obtained to reflect the regularity of the edge shape of the target superpixel block. The greater the shape regularity, the more the edge shape characteristics of the target superpixel block conform to the non-natural geological area, which provides a data basis for the subsequent accurate calculation of the non-natural feature coefficient of the target superpixel block.

[0061] Preferably, in one embodiment of the present invention, the method for acquiring the shape regularity of the target superpixel block specifically includes:

[0062] First, curve fitting is performed on the pixel points on the edge of the target superpixel block to obtain the edge fitting curve of the target superpixel block. The curve fitting method can use the existing least squares method or other methods, which are not limited here, and multiple inflection points on the edge fitting curve are obtained, wherein any two adjacent inflection points divide the edge fitting curve into a curve sub-segment. The inflection point is a point on the edge fitting curve where the second-order derivative is equal to the value 0. The inflection point is an existing concept and will not be elaborated here.

[0063] The more regular the edge shape of the target superpixel block is, the smaller the curvature of each curve subsegment of the edge fitting curve is. Therefore, any curve subsegment of the edge fitting curve can be used as the target curve subsegment, and the overall level of curvature of all pixel points on the target curve subsegment is analyzed and negatively mapped to obtain the first shape regularity parameter of the target curve subsegment, which provides a data basis for the subsequent calculation of the shape regularity of the target superpixel block. The calculation of curvature is a technical means well known to those skilled in the art and will not be elaborated here.

[0064] In the embodiment of the present invention, the overall level analysis of the curvature of all pixel points on the target curve subsegment may be achieved by calculating the average value or median of the curvature of all pixel points on the target curve subsegment, which is not limited here.

[0065] The more regular the edge shape of the target superpixel block is, the greater the difference in the extension direction of each curve subsegment is. Therefore, the overall level of the slope of all pixel points on the target curve subsegment is first analyzed to obtain the directional characteristic value of the target curve subsegment. The directional characteristic value of each curve subsegment can be obtained by the same method. The directional characteristic value reflects the characteristics of the extension direction of each curve subsegment. The calculation of the slope is a technical means well known to those skilled in the art and will not be repeated here. Then, the absolute value of the difference between the directional characteristic values ​​of the target curve subsegment and each other curve subsegment except the target curve subsegment is taken as the directional difference between the target curve subsegment and each other curve subsegment. The greater the directional difference, the greater the difference in the extension direction between the target curve subsegment and the other curve subsegments. Then, the average value of the directional difference between the target curve subsegment and all other curve subsegments is taken as the second shape regularity parameter of the target curve subsegment, providing a data basis for the subsequent calculation of the shape regularity of the target superpixel block.

[0066] In an embodiment of the present invention, the average value or median of the slopes of all pixel points on the target curve subsegment can be used as the directional characteristic value of the target curve subsegment to achieve an overall level analysis of the slopes of all pixel points on the target curve subsegment. No limitation is made here. It should be noted that pixel points on the target curve subsegment that do not have a slope do not participate in the calculation of the directional characteristic value.

[0067] Then, the first shape regularity parameter and the second shape regularity parameter are integrated to obtain the comprehensive shape regularity parameter of the target curve subsegment. The larger the comprehensive shape regularity parameter is, the more regular the shape of the target curve subsegment is. The comprehensive shape regularity parameter of each curve subsegment can be obtained by the same method, and then the average value of the comprehensive shape regularity parameters of all curve subsegments of the edge fitting curve is taken as the shape regularity of the target superpixel block.

[0068] In an embodiment of the present invention, the sum or product of the first shape regularization parameter and the second shape regularization parameter may be used as a comprehensive shape regularization parameter of the target curve subsegment to achieve a combination of the two, which is not limited here.

[0069] As an example, in one embodiment of the present invention, the expression of the shape regularity of the target superpixel block may be specifically, for example, as follows:

[0070]

[0071]

[0072]

[0073] in, Indicates the shape regularity of the target superpixel block; The first edge fitting curve of the target superpixel block The first shape regularization parameter of the curve sub-segment; The first edge fitting curve of the target superpixel block A second shape regularization parameter of a curve sub-segment; Indicates The comprehensive shape regularization parameters of the curve sub-segments; represents the number of curve subsegments of the edge fitting curve, then Indicates the number of other curve subsegments on the edge fitting curve except the target curve subsegment; represents the first shape regularization parameter of the target curve subsegment; A second shape regularization parameter representing the target curve subsegment; Indicates the first The curvature of each pixel; Indicates the number of pixels on the target curve segment; Indicates the directional characteristic value of the target curve subsegment; Indicates the first segment other than the target curve segment. Directional eigenvalues ​​of other curve subsegments; Indicates the target curve segment and the The direction difference of other curve sub-segments; Indicates the preset first adjustment parameter, which is used to prevent the denominator from being 0. The value range is In one embodiment of the present invention, Set to 0.01, The specific value of can also be set by the implementer according to the specific implementation scenario and is not limited here.

[0074] It should be noted that in other embodiments of the present invention, negative correlation mapping may be achieved through other basic mathematical operations, which will not be elaborated herein.

[0075] At the same time, compared with natural geological areas, the texture of unnatural geological areas is more regular. In the superpixel segmentation process, pixels of different colors may exist in the same superpixel block. The color distribution consistency in the superpixel blocks belonging to the unnatural geological areas is poor. The surface of the unnatural geological areas is relatively flat and the received light is relatively uniform. The degree of brightness change in the texture direction of the superpixel blocks belonging to the unnatural geological areas is small. Therefore, the color feature value of the target superpixel block can be obtained according to the distribution of the hue channel value of the pixel points in the target superpixel block and the change of the brightness channel value of the pixel points. The larger the color feature value, the more the hue and brightness characteristics inside the target superpixel block conform to the unnatural geological area, which provides a data basis for the subsequent accurate calculation of the unnatural feature coefficient of the target superpixel block.

[0076] Preferably, in one embodiment of the present invention, the method for obtaining the color feature value of the target superpixel block specifically includes:

[0077] The discrete degree of the hue channel values ​​of all pixels in the target superpixel block is analyzed to obtain the hue distribution parameter of the target superpixel block. The larger the hue distribution parameter, the poorer the consistency of the color distribution of the pixels in the target superpixel block, and the more its color distribution characteristics conform to the non-natural category geological area.

[0078] In an embodiment of the present invention, the variance or standard deviation of the hue channel values ​​of all pixels in the target superpixel block can be used as the hue distribution parameter of the target superpixel block to analyze the discrete degree of the hue channel values ​​of all pixels in the target superpixel block, which is not limited here.

[0079] Since superpixel segmentation can divide pixels with similar texture features into the same superpixel block, the texture extension direction within the superpixel block is usually consistent with the edge extension direction of the superpixel block. Therefore, first, on the edge fitting curve of the target superpixel block, the curve subsegment with the longest length or the largest number of pixels is used as the reference curve subsegment of the target superpixel block. The extension direction of the reference curve subsegment can be considered as the internal texture direction of the target superpixel block. Therefore, the inverse tangent function can be used to map the directional eigenvalues ​​of the reference curve subsegment to obtain the texture direction angle of the target superpixel block.

[0080] Then, the absolute value of the difference in brightness channel value between each pixel in the target superpixel block and the next adjacent pixel in the texture direction angle is used as the brightness change value of each pixel in the target superpixel block. The smaller the brightness change value, the smaller the pixel in the texture direction, which further indicates that the brightness change in the target superpixel block is more consistent with the non-natural category geological area. Therefore, the average value of the brightness change value of all pixels in the target superpixel block can be negatively correlated to obtain the brightness change parameter of the target superpixel block.

[0081] Then, the hue distribution parameters and brightness change parameters are integrated to obtain the color feature value of the target superpixel block.

[0082] In an embodiment of the present invention, the sum or product of the hue distribution parameter and the brightness change parameter may be used as the color feature value of the target superpixel block to achieve a combination of the two, which is not limited here.

[0083] As an example, in one embodiment of the present invention, the expression of the color feature value of the target superpixel block may be specifically, for example, as follows:

[0084]

[0085] in, Represents the color feature value of the target superpixel block; Represents the hue distribution parameter of the target superpixel block; represents the first The brightness change value of each pixel; Represents the number of pixels in the target superpixel block; Represents the brightness change parameter of the target superpixel block; Indicates the preset second adjustment parameter, which is used to prevent the denominator from being 0. The value range is In one embodiment of the present invention, Set to 0.01, The specific value of can also be set by the implementer according to the specific implementation scenario and is not limited here.

[0086] It should be noted that in other embodiments of the present invention, negative correlation mapping may be achieved through other basic mathematical operations, which will not be elaborated herein.

[0087] The larger the shape regularity and color feature value of the target superpixel block, the more the edge shape, hue and brightness characteristics of the target superpixel block conform to the non-natural category geological region. Therefore, the non-natural feature coefficient of the target superpixel block can be obtained based on the shape regularity and color feature value. The larger the non-natural feature coefficient, the more likely the target superpixel block belongs to the non-natural category geological region. Subsequently, based on the non-natural feature coefficient, the superpixel blocks of different categories from the target geological region due to noise interference can be selected from the target geological region, thereby accurately analyzing the degree of noise interference on the target geological region and improving the final image enhancement effect.

[0088] Preferably, in one embodiment of the present invention, the method for obtaining the unnatural characteristic coefficient of the target superpixel block specifically includes:

[0089] The product of shape regularity and color feature value is normalized to obtain the unnatural feature coefficient of the target superpixel block.

[0090] In one embodiment of the present invention, the normalization process may specifically be, for example, maximum and minimum value normalization process. In other embodiments of the present invention, other normalization methods may be selected according to a specific range of values, which will not be described in detail.

[0091] As an example, in one embodiment of the present invention, the expression of the unnatural feature coefficient of the target superpixel block may be specifically, for example, as follows:

[0092]

[0093] in, The coefficient representing the unnatural features of the target superpixel block; Indicates the shape regularity of the target superpixel block; Represents the color feature value of the target superpixel block; Represents the normalization function.

[0094] The unnatural feature coefficient of each superpixel block in the target geological area can be obtained by the same method as above.

[0095] Step S3: Take any geological region adjacent to the target geological region as a reference geological region, and obtain the initial interference index of the target geological region with respect to the reference geological region according to the position distribution of the centroids between the target geological region and the reference geological region, the non-natural feature coefficients of the superpixel blocks in the target geological region, the distance between the centroids of the superpixel blocks and the reference geological region, and the category of the target geological region; obtain the ground feature classification interference index of the target geological region according to the initial interference index of the target geological region with respect to all adjacent geological regions.

[0096] Considering that the segmentation result in step S1 may deviate due to the influence of noise, the target geological region usually includes part of the geological region adjacent to it. This phenomenon is more obvious at the boundary between the target geological region and the adjacent geological region. That is to say, due to the interference of noise, some superpixel blocks in the target geological region have different geological characteristics from the target geological region, but the same geological characteristics as the adjacent geological region. Therefore, in the embodiment of the present invention, any geological region adjacent to the target geological region is first used as a reference geological region, and then the position distribution of the centroid between the target geological region and the reference geological region, the non-natural feature coefficient of the superpixel block in the target geological region, the distance between the centroid of the superpixel block and the reference geological region, and the category of the target geological region are analyzed. The initial interference coefficient obtained reflects the possibility that the part of the target geological region including the adjacent geological region is caused by noise interference, which provides a data basis for the subsequent analysis of the interference of noise on the target geological region. Among them, obtaining the centroid of a certain area of ​​the image is a technical means well known to those skilled in the art and will not be elaborated here.

[0097] Preferably, in one embodiment of the present invention, the method for obtaining the initial interference index of the target geological region with respect to the reference geological region specifically includes:

[0098] See also Figure 2 , which shows a flow chart of a method for obtaining an initial interference index of a target geological region with respect to a reference geological region provided by an embodiment of the present invention.

[0099] Step S301: Connect the centroid of the target geological area and the centroid of the reference geological area with a straight line to obtain a centroid line between the target geological area and the reference geological area, and use a straight line passing through the centroid of the target geological area and perpendicular to the centroid line as a dividing line of the target geological area. The dividing line divides the target geological area into two sub-areas, and uses the sub-area adjacent to the reference geological area as the target sub-area of ​​the target geological area.

[0100] See also Figure 3, which shows a schematic diagram of a target sub-region of a target geological region provided by an embodiment of the present invention, wherein point X is the centroid of the target geological region, point Y is the centroid of the reference geological region, L1 is a centroid line between the target geological region and the reference geological region, L2 is a dividing line of the target geological region, regions Q1 and Q2 are two sub-regions of the target geological region, and sub-region Q1 is a target sub-region of the target geological region.

[0101] Step S302: taking the ordered sequence composed of the non-natural feature coefficients of all superpixel blocks in the target sub-region as a reference sequence, and obtaining the segmentation threshold of the target geological region according to the difference between two adjacent non-natural feature coefficients in the reference sequence.

[0102] Due to the interference of noise, there are superpixel blocks in the target geological area that do not belong to the category to which the target geological area belongs, and the non-natural feature coefficients of these superpixel blocks are quite different from those of other superpixel blocks in the target geological area. Therefore, in an embodiment of the present invention, an ordered sequence composed of the non-natural feature coefficients of all superpixel blocks in the target sub-area is first used as a reference sequence. The order of the non-natural feature coefficients in the reference sequence can be increasing or decreasing, and then the segmentation threshold of the target geological area is obtained based on the difference between two adjacent non-natural feature coefficients in the reference sequence. The segmentation threshold can be used to accurately select superpixel blocks of different categories in the target geological area.

[0103] Preferably, in one embodiment of the present invention, the method for obtaining the segmentation threshold of the target geological area specifically includes:

[0104] Any two adjacent non-natural feature coefficients in the reference sequence are taken as an element group, and the absolute value of the difference between the two non-natural feature coefficients in each element group is taken as the element change amount of each element group. The larger the element change amount, the greater the degree of mutation of the non-natural feature coefficient of the superpixel block in the target sub-region in the element group, which further indicates that the element group is more likely to be the dividing point for distinguishing superpixel blocks of different categories. Therefore, the average value of the two non-natural feature coefficients in the element group corresponding to the maximum value of the element change amount can be used as the segmentation threshold of the target geological area.

[0105] Step S303: selecting superpixel blocks of different categories in the target geological area from all superpixel blocks in the target geological area according to the non-natural feature coefficient of each superpixel block in the target geological area, the category of the target geological area and the segmentation threshold.

[0106] The larger the non-natural feature coefficient, the more the superpixel block conforms to the non-natural category geological region. On the contrary, the smaller the non-natural feature coefficient, the more the superpixel block conforms to the natural category geological region. The segmentation threshold is the dividing value for distinguishing superpixel blocks of different categories in the target geological region. Therefore, according to the non-natural feature coefficient and segmentation threshold of each superpixel block in the target geological region, the non-same category superpixel blocks of the target geological region can be accurately selected from all superpixel blocks in the target geological region in combination with the category of the target geological region. Subsequently, the non-same category superpixel blocks of the target geological region can be analyzed to accurately calculate the initial interference index of the target geological region with respect to the reference geological region.

[0107] Preferably, in one embodiment of the present invention, the method for obtaining superpixel blocks of different categories specifically includes:

[0108] If the category of the target geological area is a natural category, the larger the non-natural feature coefficient of the superpixel block, the more the superpixel block conforms to the non-natural category geological area. Therefore, in the target geological area, the superpixel block corresponding to the non-natural feature coefficient greater than the segmentation threshold can be selected as the non-same category superpixel block of the target geological area.

[0109] If the category of the target geological area is a non-natural category, the smaller the non-natural feature coefficient of the superpixel block, the more the superpixel block conforms to the natural category geological area. In the target geological area, the superpixel block corresponding to the non-natural feature coefficient less than the segmentation threshold is selected as the non-same category superpixel block of the target geological area.

[0110] Step S304: the absolute value of the difference between the non-natural feature coefficient of each non-same superpixel block and the segmentation threshold of the target geological area is used as the first anomaly coefficient of each non-same superpixel block; the Euclidean distance between the centroid of each non-same superpixel block and the centroid of the reference geological area is negatively correlated to obtain the second anomaly coefficient of each non-same superpixel block; the first anomaly coefficient and the second anomaly coefficient are combined to obtain the combined anomaly coefficient of each non-same superpixel block.

[0111] The greater the difference between the unnatural feature coefficient of non-same superpixel blocks and the segmentation threshold, the more likely that the abnormal phenomenon of non-same superpixel blocks appearing in the target geological area is caused by noise interference. Therefore, the absolute value of the difference between the unnatural feature coefficient of each non-same superpixel block and the segmentation threshold of the target geological area can be used as the first abnormal coefficient of each non-same superpixel block. Moreover, the closer the distance between the centroid of the non-same superpixel block and the centroid of the reference geological area, the more likely that the non-same superpixel block is the same as the geological characteristics of the reference geological area, which further indicates that the abnormal phenomenon of non-same superpixel blocks appearing in the target geological area is more likely to be caused by noise interference. Therefore, the Euclidean distance between the centroid of each non-same superpixel block and the centroid of the reference geological area can be negatively correlated to obtain the second abnormal coefficient of each non-same superpixel block, and then the first abnormal coefficient and the second abnormal coefficient can be combined to obtain the comprehensive abnormal coefficient of each non-same superpixel block.

[0112] In an embodiment of the present invention, the sum or product of the first abnormal coefficient and the second abnormal coefficient of each non-same category superpixel block can be used as the comprehensive abnormal coefficient of each non-same category superpixel block to achieve the integration of the two, which is not limited here.

[0113] Step S305: taking the accumulated value of the comprehensive abnormal coefficients of all superpixel blocks of different categories in the target geological region as the initial interference index of the target geological region with respect to the reference geological region.

[0114] The larger the comprehensive anomaly coefficient is, the more likely it is that the abnormal phenomenon of superpixel blocks of different categories appearing in the target geological area relative to the reference geological area is caused by noise interference. Therefore, the accumulated value of the comprehensive anomaly coefficients of all superpixel blocks of different categories in the target geological area can be used as the initial interference index of the target geological area with respect to the reference geological area.

[0115] As an example, in one embodiment of the present invention, the expression of the initial interference index of the target geological region with respect to the reference geological region may be specifically, for example, as follows:

[0116]

[0117] in, represents the initial disturbance index of the target geological area with respect to the reference geological area; Indicates the first Unnatural characteristic coefficients of superpixel blocks of different categories; represents the segmentation threshold of the target geological area; Indicates The Euclidean distance between the centroid of the superpixel blocks of different categories and the centroid of the reference geological area; Indicates The first abnormal coefficient of superpixel blocks of different categories; Indicates The second abnormal coefficient of superpixel blocks of different categories; Indicates the number of superpixel blocks of different categories in the target geological area; Indicates The comprehensive abnormal coefficient of superpixel blocks of different categories; Indicates the preset third adjustment parameter, which is used to prevent the denominator from being 0. The value range is In one embodiment of the present invention, Set to 0.01, The specific value of can also be set by the implementer according to the specific implementation scenario and is not limited here.

[0118] The same method as above can be used to obtain the initial interference index of the target geological area with respect to each adjacent geological area. The larger the initial interference index, the more likely it is that part of the target geological area containing adjacent geological areas is caused by noise interference. Therefore, the ground object classification interference index of the target geological area can be obtained based on the initial interference index of the target geological area with respect to all adjacent geological areas. The larger the ground object classification interference index, the greater the degree of noise interference to the target geological area. Subsequently, the multispectral image can be effectively enhanced based on the ground object classification interference index.

[0119] Preferably, in one embodiment of the present invention, the average value of the initial interference indexes of the target geological region with respect to all adjacent geological regions is used as the ground feature classification interference index of the target geological region.

[0120] As an example, in one embodiment of the present invention, the expression of the object classification interference index of the target geological area can be specifically, for example, as follows:

[0121]

[0122] in, Indicates the interference index of the land object classification in the target geological area; Indicates the target geological area with respect to Initial disturbance index of adjacent geological areas; Indicates the number of geological areas adjacent to the target geological area.

[0123] The ground object classification interference index of each geological area in the multispectral image can be obtained by the same method as above.

[0124] Step S4: Based on the ground object classification interference index of each geological area, the multispectral image is denoised to obtain an enhanced image.

[0125] The larger the ground object classification interference index of each geological area, the more serious the phenomenon of the appearance of partial areas with different characteristics from the geological area in the geological area, which further indicates that the degree of interference of noise on the geological area is greater. Therefore, based on the ground object classification interference index of each geological area, the multispectral image can be denoised to obtain an enhanced image. Compared with the existing use of fixed preset denoising parameters, denoising and enhancement can be performed based on the degree of noise interference on different geological areas, thereby improving the image enhancement effect.

[0126] Preferably, in one embodiment of the present invention, the method for acquiring an enhanced image specifically includes:

[0127] The ground object classification interference index of each geological area is used as a smoothing parameter used in a filtering denoising algorithm, and the filtering denoising algorithm is used to denoise each geological area in the multispectral image to obtain an enhanced image. For example, the ground object classification interference index of each geological area can be used as a standard deviation parameter used in a Gaussian filtering algorithm, and the Gaussian filtering algorithm is used to perform filtering denoising on the corresponding geological area to obtain an enhanced image.

[0128] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0129] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A method for enhancing geological surveying and remote sensing images, characterized in that: The method comprises: Acquire a multispectral image of the geological survey area to be measured, perform color space conversion on the multispectral image, and obtain a hue channel value and a brightness channel value of each pixel in the multispectral image; segment the multispectral image to obtain different geological areas in the multispectral image and a category of each geological area, wherein the category includes a natural category and a non-natural category; Taking any geological region as a target geological region, performing superpixel segmentation on the target geological region to obtain multiple superpixel blocks of the target geological region, taking any superpixel block in the target geological region as a target superpixel block, and obtaining the shape regularity of the target superpixel block according to the direction of the pixel points on the edge of the target superpixel block; obtaining the color feature value of the target superpixel block according to the distribution of the hue channel value of the pixel points in the target superpixel block and the change of the brightness channel value of the pixel points; obtaining the non-natural feature coefficient of the target superpixel block based on the shape regularity and the color feature value; Taking any geological region adjacent to the target geological region as a reference geological region, obtaining an initial interference index of the target geological region with respect to the reference geological region according to the position distribution of the centroids between the target geological region and the reference geological region, the non-natural feature coefficients of the superpixel blocks in the target geological region, the distance between the centroids of the superpixel blocks and the reference geological region, and the category of the target geological region; obtaining a ground feature classification interference index of the target geological region according to the initial interference index of the target geological region with respect to all adjacent geological regions; Based on the object classification interference index of each geological region, denoising the multispectral image to obtain an enhanced image; The obtaining of the initial interference index of the target geological region with respect to the reference geological region comprises: Connecting the centroid of the target geological region and the centroid of the reference geological region with a straight line to obtain a centroid line between the target geological region and the reference geological region, and taking a straight line passing through the centroid of the target geological region and perpendicular to the centroid line as a segmentation line of the target geological region, the segmentation line divides the target geological region into two sub-regions, and taking the sub-region adjacent to the reference geological region as a target sub-region of the target geological region; Taking the ordered sequence composed of the non-natural feature coefficients of all superpixel blocks in the target sub-region as a reference sequence, and obtaining the segmentation threshold of the target geological region according to the difference between two adjacent non-natural feature coefficients in the reference sequence; According to the non-natural feature coefficient of each super pixel block in the target geological region, the category of the target geological region and the segmentation threshold, super pixel blocks of different categories in the target geological region are selected from all super pixel blocks in the target geological region; The absolute value of the difference between the non-natural feature coefficient of each non-same superpixel block and the segmentation threshold of the target geological area is used as the first abnormal coefficient of each non-same superpixel block; the Euclidean distance between the centroid of each non-same superpixel block and the centroid of the reference geological area is negatively correlated to obtain the second abnormal coefficient of each non-same superpixel block; the first abnormal coefficient and the second abnormal coefficient are combined to obtain the comprehensive abnormal coefficient of each non-same superpixel block; The accumulated value of the comprehensive anomaly coefficients of all superpixel blocks of different categories in the target geological region is used as the initial interference index of the target geological region with respect to the reference geological region; The step of obtaining the segmentation threshold of the target geological area according to the difference between two adjacent non-natural feature coefficients in the reference sequence comprises: Any two adjacent non-natural characteristic coefficients in the reference sequence are regarded as an element group, and the absolute value of the difference between the two non-natural characteristic coefficients in each element group is regarded as the element variation of each element group; The average value of two non-natural characteristic coefficients in the element group corresponding to the maximum value of the element variation is used as the segmentation threshold of the target geological area; The step of selecting superpixel blocks of different categories in the target geological area from all superpixel blocks in the target geological area comprises: If the category of the target geological area is a natural category, then in the target geological area, a superpixel block corresponding to a non-natural feature coefficient greater than a segmentation threshold is selected as a superpixel block of a non-same category in the target geological area; If the category of the target geological area is a non-natural category, then in the target geological area, a superpixel block corresponding to a non-natural feature coefficient smaller than the segmentation threshold is selected as a non-same category superpixel block of the target geological area; The obtaining of the shape regularity of the target superpixel block comprises: Performing curve fitting on the pixel points on the edge of the target super-pixel block to obtain the edge fitting curve of the target super-pixel block; Acquire a plurality of inflection points on the edge fitting curve, wherein any two adjacent inflection points divide the edge fitting curve into a curve subsegment; Taking any curve subsegment of the edge fitting curve as the target curve subsegment, analyzing the overall level of curvature of all pixel points on the target curve subsegment and performing negative correlation mapping to obtain the first shape regularization parameter of the target curve subsegment; Analyze the overall level of the slopes of all pixel points on the target curve subsegment to obtain the directional characteristic value of the target curve subsegment; use the absolute value of the difference between the directional characteristic value of the target curve subsegment and each other curve subsegment except the target curve subsegment as the directional difference between the target curve subsegment and each other curve subsegment; use the average value of the directional difference between the target curve subsegment and all other curve subsegments as the second shape regularization parameter of the target curve subsegment; The first shape regularization parameter and the second shape regularization parameter are integrated to obtain an integrated shape regularization parameter of the target curve subsegment; Taking the average value of the comprehensive shape regularity parameters of all curve subsegments of the edge fitting curve as the shape regularity of the target superpixel block; The step of obtaining the color feature value of the target super pixel block comprises: Analyze the discrete degree of the hue channel values ​​of all pixels in the target superpixel block to obtain the hue distribution parameter of the target superpixel block; On the edge fitting curve of the target superpixel block, the curve subsegment with the largest length is used as the reference curve subsegment of the target superpixel block, and the directional feature value of the reference curve subsegment is mapped by using an inverse tangent function to obtain the texture direction angle of the target superpixel block; The absolute value of the difference between the brightness channel value of each pixel in the target super-pixel block and the next adjacent pixel at the texture direction angle is used as the brightness change value of each pixel in the target super-pixel block; the average value of the brightness change values ​​of all pixels in the target super-pixel block is negatively correlated to obtain the brightness change parameter of the target super-pixel block; The hue distribution parameter and the brightness change parameter are integrated to obtain a color feature value of the target superpixel block; The obtaining of the unnatural characteristic coefficient of the target superpixel block based on the shape regularity and the color characteristic value comprises: Normalizing the product of the shape regularity and the color feature value to obtain a non-natural feature coefficient of the target superpixel block; The obtaining of the interference index of the ground object classification of the target geological area comprises: The average value of the initial interference indexes of the target geological region with respect to all adjacent geological regions is used as the ground feature classification interference index of the target geological region.

2. A geological surveying and remote sensing image enhancement method according to claim 1, characterized in that: Denoising the multispectral image based on the ground object classification interference index of each geological region to obtain an enhanced image includes: The ground object classification interference index of each geological region is used as a smoothing parameter used by a filtering denoising algorithm, and the filtering denoising algorithm is used to denoise each geological region in the multispectral image to obtain an enhanced image.

3. A geological surveying and remote sensing image enhancement method according to claim 1, characterized in that: The obtaining of different geological regions and categories of each geological region in the multispectral image includes: Acquire a plurality of multispectral image samples from a database, wherein the multispectral image samples contain regions with different annotated categories, and the annotated categories are divided into natural categories and non-natural categories; The support vector machine is trained using multispectral image samples, and the trained support vector machine is used to segment the multispectral image to obtain different geological areas in the multispectral image and the category of each geological area.

Images