A method for correction and enhancement of land remote sensing images

Abstract

The present invention relates to the technical field of remote sensing image enhancement, and in particular to a land remote sensing image correction and enhancement method, the method comprising: collecting ground land fog remote sensing images; obtaining each region of the land fog remote sensing images; obtaining the regional internal steepness index of the region according to the grayscale distribution of the pixels in the region; constructing the local steepness heterogeneity coefficient of each pixel in the region; constructing the sort pixel set and high heterogeneity pixel sequence of each region, and then obtaining the local steepness heterogeneity distance sequence; constructing the distance sequence disorder index according to the element mean, maximum value and minimum value of the local steepness heterogeneity distance sequence; obtaining the adaptive filtering window size of each region according to the distance sequence disorder index; using the dark channel algorithm and combining the adaptive filtering window size of each region to complete the adaptive enhancement correction of the land fog remote sensing image. The present invention can realize the enhancement processing of the land remote sensing image and ensure the reliability of the land remote sensing image information.

Description

Technical Field

[0001] The present application relates to the technical field of remote sensing image enhancement, and in particular to a method for correcting and enhancing a land remote sensing image. Background Art

[0002] After decades of development, optical remote sensing technology has been widely used in people's daily lives, especially in ecological protection, urban planning, agricultural management, disaster monitoring, and national defense and military fields. In the process of collecting remote sensing images, it is necessary to penetrate the atmosphere to shoot the ground. However, the influence of aerosols such as clouds and fog has brought difficulties to the analysis and subsequent work of remote sensing images. In order to overcome this problem, remote sensing image enhancement technology can effectively remove clouds and fog in images. Obtaining high-quality and high-definition remote sensing images is a prerequisite for carrying out the above-mentioned related research. Therefore, it is of great significance to enhance land remote sensing images, which is conducive to improving the available information and information extraction capabilities of remote sensing images.

[0003] The dark channel prior algorithm is a simple and effective image enhancement technology. However, the traditional dark channel defogging algorithm uses a fixed window to obtain the dark channel image. The optimal window size required in different scenes is also different. A single window is prone to color image distortion, and the enhanced image has halos, and the image effect is not good in the case of large amounts of fog. Since the building area and vegetation area account for the largest proportion in urban remote sensing images, there is also a defect that when the roof appears green, it is easy to be distinguished as a vegetation area.

[0004] In summary, the present invention proposes a land remote sensing image correction and enhancement method, which collects land cloud and fog remote sensing images, performs preprocessing, divides the image area, sets an adaptive local window size for obtaining a dark channel image according to the degree of cloud and fog that needs to be eliminated in each divided area, uses a filter window of adaptive size to obtain a dark channel image, and uses a dark channel prior algorithm to enhance the remote sensing image. Summary of the invention

[0005] In order to solve the above technical problems, the present invention provides a land remote sensing image correction and enhancement method to solve the existing problems.

[0006] A land remote sensing image correction and enhancement method of the present invention adopts the following technical solution:

[0007] An embodiment of the present invention provides a land remote sensing image correction and enhancement method, the method comprising the following steps:

[0008] Collect ground land cloud remote sensing images and perform preprocessing;

[0009] Obtain each region of the land cloud remote sensing image; for each region, obtain the regional internal steepness index of the region according to the grayscale distribution of the pixel points in the region; construct the local steepness heterogeneity coefficient of each pixel point in the region according to the difference between the window internal steepness index of each sliding window in the region and the regional internal steepness index; construct the sort pixel set of each region according to the local steepness heterogeneity coefficient of each pixel point in the region; obtain the high heterogeneity pixel sequence by combining the distance relationship between each element in the sort pixel set of each region; obtain the local steepness heterogeneity distance sequence according to the distance between each element in the high heterogeneity pixel sequence; construct the distance sequence disorder index according to the element mean, maximum value and minimum value of the local steepness heterogeneity distance sequence; obtain the adaptive filtering window size of each region according to the distance sequence disorder index;

[0010] The dark channel algorithm is used in combination with the adaptive filtering window size of each area to complete the adaptive enhancement correction of land cloud remote sensing images.

[0011] Furthermore, the acquisition of land cloud and fog remote sensing images of various regions includes:

[0012] The grayscale image of the land cloud remote sensing image is obtained, and the canny edge detection algorithm is used to detect the edge of the grayscale image, and a binary image is output. The edge pixels divide the land cloud remote sensing image into various regions.

[0013] Furthermore, the acquisition of land cloud and fog remote sensing images of various regions includes:

[0014] Obtain a grayscale image of the land cloud remote sensing image, and use the grayscale image as the input of the semantic segmentation network. The output of the semantic segmentation network is a semantic segmentation rendering, where the semantic segmentation rendering is a binary image that divides the land cloud remote sensing image into various regions.

[0015] Furthermore, obtaining the internal steepness index of the region according to the grayscale distribution of pixels in the region includes:

[0016] Calculate the grayscale mean and variance of the pixels in the area, obtain the difference between the grayscale value of each pixel in the area and the grayscale mean, calculate the square of the ratio of the difference to the variance, and take the mean of the squares of all pixels in the area as the internal steepness index of the area.

[0017] Furthermore, the method of constructing the local steepness heterogeneity coefficient of each pixel point in the region according to the difference between the steepness index inside the window of each sliding window in the region and the steepness index inside the region includes:

[0018] A local window of w*w is constructed with each pixel in the region as the center, where w is the side length of the local window. The internal steepness index of the window in the local window is calculated, and the absolute value of the difference between the internal steepness index of the window and the internal steepness index of the region is obtained. The absolute value of the difference is used as the local steepness heterogeneity coefficient of the central pixel of the local window.

[0019] Furthermore, constructing the sort pixel set of each region according to the local steepness heterogeneity coefficient of each pixel point in the region includes:

[0020] For each region;

[0021] The local steepness heterogeneity coefficients of each pixel point in the region are arranged in descending order from large to small, and the pixels corresponding to the first W% of the local steepness heterogeneity coefficients are taken to form the sort pixel set.

[0022] Furthermore, the step of combining the distance relationship between the elements in the sort pixel set of each region to obtain a highly heterogeneous pixel sequence includes:

[0023] For the sort pixel set of each region;

[0024] Randomly select any pixel point in the sort pixel set as a random point, calculate the Manhattan distance between other pixel points in the sort pixel set and the random point, and use the pixel point with the smallest Manhattan distance to the random point as a new random point. Similarly, traverse the sort pixel set to obtain each random point, and compose all random points into a highly heterogeneous pixel sequence.

[0025] Furthermore, the obtaining of a local steep heterogeneity distance sequence according to the distance between each element in the high heterogeneity pixel sequence includes:

[0026] The Manhattan distances between adjacent elements in the high heterogeneity pixel sequence are obtained, and the Manhattan distances between all adjacent elements in the high heterogeneity pixel sequence are used as the elements of the local steep heterogeneity distance sequence.

[0027] Furthermore, constructing a distance sequence disorder index according to the element mean, maximum value and minimum value of the local steep heterogeneity distance sequence includes:

[0028] Get the element mean, maximum and minimum value of the local steep heterogeneity distance sequence;

[0029] Calculate the square of the difference between each element of the local steep heterogeneity distance sequence and the mean of the element, obtain the mean of the square of the difference of all elements of the local steep heterogeneity distance sequence and then take the square root; obtain the difference between the maximum value and the minimum value, and use the difference as the negative exponent of an exponential function with a natural constant as the base;

[0030] The ratio of the square root to the calculation result of the exponential function is used as the distance sequence disorder index.

[0031] Furthermore, the step of obtaining the adaptive filtering window size of each region according to the distance sequence disorder index includes:

[0032] Obtaining a ratio of a distance sequence disorder index to a preset adjustment parameter, obtaining a calculation result of an exponential function with the ratio as an exponent and a natural constant as a base, recording the result as a first result, and calculating a difference between 1 and the first result;

[0033] Preset a partition coefficient and a basic window coefficient for each region, obtain the product of the partition coefficient and the difference, and use the sum of the basic window coefficient and the product as the adaptive filtering window size of the region.

[0034] The present invention has at least the following beneficial effects:

[0035] The present invention constructs the local steep heterogeneity coefficient in the region by analyzing the pixel value variation characteristics of the vegetation area and the building area, further obtains the local steep heterogeneity distance sequence through the coefficient, analyzes the fluctuation level of the sequence to obtain the distance sequence disorder index, and uses it to accurately identify the building area and the vegetation area. According to the characteristics of the two areas, the adaptive dark channel image extraction window size is set, which makes up for the image enhancement distortion and halo caused by the single fixed window used in the existing method, and also avoids the defect of identifying the building area as the vegetation area. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] 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.

[0037] Figure 1 A flow chart of a land remote sensing image correction and enhancement method provided by the present invention;

[0038] Figure 2 Flowchart for setting adaptive filter window size for each region;

[0039] Figure 3 This is a schematic diagram comparing the land cloud remote sensing image and the enhanced high-definition image. DETAILED DESCRIPTION

[0040] 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 land remote sensing image correction and 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.

[0041] 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.

[0042] The specific scheme of the land remote sensing image correction and enhancement method provided by the present invention is described in detail below with reference to the accompanying drawings.

[0043] An embodiment of the present invention provides a land remote sensing image correction and enhancement method. Specifically, the following land remote sensing image correction and enhancement method is provided. Figure 1 , the method comprises the following steps:

[0044] Step S001, obtaining land cloud and fog remote sensing images and preprocessing the images.

[0045] The RGB remote sensing image of the ground is obtained by an optical sensor, which is recorded as a land fog remote sensing image in this embodiment. Noise may be generated during the shooting process due to factors such as sensor equipment and atmospheric interference, which will affect the subsequent analysis, so the image is preprocessed first. The image is denoised by using the Gaussian filtering method, that is, the Gaussian function is convolved with the acquired fog image to eliminate random noise. It should be noted that the Gaussian filtering denoising process is an existing well-known technology and will not be elaborated in detail here.

[0046] At this point, the RGB remote sensing image of the ground can be obtained as the basic image data for land remote sensing image enhancement and correction.

[0047] Step S002: Divide the image area, and set the local window size for obtaining the dark channel image adaptively according to the degree of fog that needs to be eliminated in each divided area.

[0048] The extraction of dark channel images in the original algorithm is divided into two steps: Since the collected cloud remote sensing color images have three channels, RGB, for a certain pixel, the smallest value among the three channels R, G, and B is taken as the grayscale value of the pixel. This operation is performed on each pixel in the image to obtain the initial dark channel image, recorded as dark1; use The local window of performs minimum filtering on each point, that is, the minimum gray value in the window is selected to replace the central pixel, thereby obtaining the dark channel image.

[0049] When calculating the transmittance image based on this principle, using different filter windows will result in different final results. For example, the smaller the filter window, the better the effect of removing clouds and fog, and the larger the filter window, the better the effect of retaining the color and details in the image. There are water surface areas, vegetation areas and building areas in the land cloud remote sensing image, and different locations require different degrees of cloud removal. In the subsequent analysis of the enhanced image, the accuracy of the building area and vegetation area in the image is required to be higher, and the accuracy requirement for the water surface area is lower. Therefore, the adaptive window size is set according to different areas to extract the dark channel image.

[0050] Identify the boundaries of interest. In urban remote sensing images, there are mainly vegetation areas and building areas, and there are obvious differences in pixel values ​​between different areas, so it is easy to extract boundary images. Convert the image from RGB image to grayscale image and obtain each area in the image.

[0051] The first embodiment of obtaining each region in the image is: using Canny edge detection on the grayscale image to obtain the boundary of each region, and outputting it as a binary image. Since the boundary lines are not necessarily continuous, the closing operation in the morphological operation is used to connect the edges of the binary image. The connected boundaries are mapped to the original image, and the land cloud remote sensing image is divided into different regions.

[0052] Another embodiment of acquiring each area in the image is: using a semantic segmentation network to segment the grayscale image, the input of the semantic segmentation network is the grayscale image, the output of the semantic segmentation network is a semantic segmentation rendering, and the semantic segmentation rendering is a binary image. In this embodiment, the grayscale value of non-ground pixel points is 1, and the grayscale value of bare ground pixel points is 0, and the land cloud remote sensing image is divided into different areas.

[0053] (1) After obtaining different areas, the next step is to identify them. Compared with the building area, the pixel value variation in the vegetation area is smaller, and after image enhancement, the building area is often the focus of analysis and requires higher clarity. Using the variation characteristics of the pixel gray value helps to distinguish vegetation from buildings and construct the steepness index inside the region:

[0054]

[0055] In the formula represents the steepness index inside the region, is the number of pixels in the area, Represents the gray value of the i-th pixel in the area, and Respectively represent the grayscale mean and variance of the pixels in the region. If the grayscale value variation of the pixels in a region is smaller, the values ​​of each point are closer to the mean. The smaller the value, the smaller the calculated steepness index inside the region. The steepness index inside the region is conducive to the subsequent accurate identification of vegetation and building parts.

[0056] (2) Furthermore, the building area contains more details than the vegetation area. When extracting dark channel images, a larger filter window size is required. Therefore, the vegetation area and the building area are identified based on their different texture distributions. The vegetation area is the part with dense trees, where the size is The texture distribution similarity between the local window area and the overall vegetation area is relatively high, while the building area has irregularities of buildings, and the similarity between the local area and the overall building area is relatively low. The value of w can be set by the implementer. In this embodiment, w=5 is set. Thus, the local steepness heterogeneity coefficient of the pixels in the area is constructed:

[0057]

[0058] In the formula represents the local steepness heterogeneity coefficient of pixel j in the region, It represents the internal steepness index of the local window with size w centered at pixel point j in the region. The calculation method of the internal steepness index of the window is the same as the calculation method of the internal steepness index of the region mentioned above, which will not be repeated here. Represents the steepness index of the interior of the region.

[0059] If the difference between the internal steepness index value of the local window in the vegetation area or the building area and the internal steepness index value of the whole area is smaller, the calculated The smaller the value, the smaller the local steepness heterogeneity coefficient of the corresponding pixel point j, the higher the texture regularity in the local window of the point, and the more likely it is to be in the vegetation area. The larger it is, the less similar the local texture distribution of the point is to the overall texture, and the more likely it is to be in a built-up area.

[0060] (3) At this point, the above method of this embodiment is repeated to calculate the local steepness heterogeneity coefficient of each pixel point in different areas. To a certain extent, it shows the richness of the pixel values ​​in the local neighborhood of the pixel point. The larger the local steepness heterogeneity coefficient, the higher the pixel value richness of the local neighborhood of the point. In the building area, the points with large local steepness heterogeneity coefficient are mostly concentrated at the edge of the building. In the vegetation area, the points with large local steepness heterogeneity coefficient are relatively discretely distributed inside or between trees.

[0061] Since each pixel corresponds to a Value, on the contrary, each Each value corresponds to a pixel point. For each area, the local steepness heterogeneity coefficients are arranged in descending order from large to small, and the pixel point sets corresponding to the top 10% of the local steepness heterogeneity coefficients are taken to form the sort pixel set. As analyzed above, the pixel point distribution of the sort pixel set is different in different areas. According to the pixel point spacing in the sort pixel set, a high-level heterogeneity pixel sequence can be constructed. The construction process of the high-level heterogeneity pixel sequence is as follows:

[0062] For the sort pixel set of each area, randomly select any pixel point in the sort pixel set as a random point, calculate the Manhattan distance between other pixel points in the sort pixel set and the random point, and use the pixel point with the smallest Manhattan distance to the random point as a new random point. Similarly, traverse the sort pixel set to obtain each random point, and form all random points into a highly heterogeneous pixel sequence. In this embodiment, an example is given as follows: randomly select a point in the sort pixel set, denoted as ,calculate The Manhattan distance between each point remaining in the sorted pixel set is The smallest point is , and then in addition and Find the distance from the point outside The closest point is , and so on, until all points in the sorted pixel set are found, and the high heterogeneity pixel sequence is recorded :

[0063]

[0064] In a high-level heterogeneity pixel sequence, and The Manhattan distance between ,Will and The Manhattan distance between , for and The Manhattan distance between the two points is calculated by analogy, and the Manhattan distance between all adjacent points in the high heterogeneity pixel sequence is obtained to obtain the local steep heterogeneity distance sequence. for:

[0065]

[0066] (4) In the building area, the pixels in the sort pixel set are distributed near the edge of the building, with a compact and regular distance. Therefore, the calculated local steep heterogeneity distance sequence values ​​are relatively close. In the vegetation area, the pixels in the sort pixel set are relatively scattered and far apart, resulting in a large calculated local steep heterogeneity distance sequence value. The distance sequence disorder index is constructed:

[0067]

[0068] In the formula, is the distance sequence disorder index, represents the total number of local steep heterogeneity distance sequences, The first order of the local steep heterogeneity distance sequence values, is the mean of the local steep heterogeneity distance sequence, exp represents the exponential function, which can avoid the situation where the denominator is zero. and They represent the minimum and maximum values ​​in the local steep heterogeneity distance sequence, respectively.

[0069] If the values ​​of the local steep heterogeneity distance sequence are closer, the difference between them and the mean is smaller. The smaller the value, the closer the maximum value is to the minimum value, making The larger it is, the smaller the corresponding distance sequence disorder index is, and the more the area belongs to the built-up area.

[0070] Different areas require different levels of cloud removal. In the process of image enhancement, the detail preservation effect of the building area is required to be better, while the defogging effect of the vegetation area is required to be better. Therefore, the filter window size used to obtain the dark channel image can be adaptively set according to the distance sequence disorder index of each area. The calculation formula is as follows:

[0071]

[0072] In the formula represents the adaptive filter window side length, is the rounding function, is the basic window coefficient, and the empirical value is 10. is the partition coefficient, which is used to accurately adjust the filter window size of the building area and the vegetation area. The empirical value is 5. is the distance sequence disorder index, It is the adjustment parameter of the distance sequence disorder index, which is used to slow down the growth rate of the exponential function value, and the empirical value is 6. The flow chart of setting the adaptive filter window size for each region is as follows: Figure 2 shown.

[0073] When the distance sequence disorder index is smaller, and and If the values ​​of are all positive, then The larger the value of is, the value range of this part is , making the final calculated dark channel image acquisition window larger.

[0074] The smaller the filter window, the better the effect of removing fog and mist, and the larger the filter window, the better the effect of retaining the color and details in the image. There are more details in the building area, and accurate edge information needs to be guaranteed for subsequent analysis, so the corresponding window side length is relatively large; the vegetation area is affected by fog and mist, and it is necessary to ensure relatively accurate detail information while having a certain defogging effect, so the window side length is relatively small.

[0075] Step S003, using a filter window of adaptive size to obtain a dark channel image, and using a dark channel priori algorithm to enhance the remote sensing image.

[0076] At this point, the adaptive filter window size is obtained through the above process, and then the adaptive window size is used to extract the dark channel image of the collected image, thereby obtaining the global atmospheric light value and transmittance image, and finally the enhanced remote sensing image is calculated based on the dark channel prior model. Figure 3 As shown, it should be noted that the image enhancement process of the dark channel prior model can be implemented through existing technologies, and the specific process will not be described in detail here.

[0077] In summary, the embodiment of the present invention constructs a local steep heterogeneity coefficient in the region by analyzing the pixel value change characteristics of the vegetation area and the building area, further obtains the local steep heterogeneity distance sequence through the coefficient, analyzes the fluctuation level of the sequence to obtain the distance sequence disorder index, and uses it to accurately identify the building area and the vegetation area; sets an adaptive dark channel image extraction window size according to the characteristics of the two areas, which makes up for the image enhancement distortion and halo caused by the single fixed window used in the existing method, and also avoids the defect of identifying the building area as the vegetation area, and has the beneficial effects of good image enhancement effect, strong adaptability, and improved reliability of remote sensing image information.

[0078] 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 above is a description of a specific embodiment of this specification. In addition, 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.

[0079] 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.

[0080] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Modifications to the technical solutions recorded in the aforementioned embodiments, or equivalent replacement of some of the technical features therein, do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A land remote sensing image correction and enhancement method, characterized in that: The method comprises the following steps: Collect ground land cloud remote sensing images and perform preprocessing; Obtain each region of the land cloud remote sensing image; for each region, obtain the regional internal steepness index of the region according to the grayscale distribution of the pixel points in the region; construct the local steepness heterogeneity coefficient of each pixel point in the region according to the difference between the window internal steepness index of each sliding window in the region and the regional internal steepness index; construct the sort pixel set of each region according to the local steepness heterogeneity coefficient of each pixel point in the region; obtain the high heterogeneity pixel sequence by combining the distance relationship between each element in the sort pixel set of each region; obtain the local steepness heterogeneity distance sequence according to the distance between each element in the high heterogeneity pixel sequence; construct the distance sequence disorder index according to the element mean, maximum value and minimum value of the local steepness heterogeneity distance sequence; obtain the adaptive filtering window size of each region according to the distance sequence disorder index; Adopting dark channel algorithm and combining adaptive filter window size of each area, adaptive enhancement correction of land cloud remote sensing image is completed; The step of obtaining the internal steepness index of the region according to the grayscale distribution of the pixels in the region includes: Calculate the grayscale mean and variance of the pixels in the region, obtain the difference between the grayscale value of each pixel in the region and the grayscale mean, calculate the square of the ratio of the difference to the variance, and take the mean of the squares of all the pixels in the region as the internal steepness index of the region; The method of constructing a sort pixel set of each region according to the local steepness heterogeneity coefficient of each pixel point in the region includes: For each region; Arrange the local steepness heterogeneity coefficients of each pixel point in the region in descending order from large to small, and take the pixels corresponding to the first W% of the local steepness heterogeneity coefficients to form a sort pixel set; The method of constructing a distance sequence disorder index according to the element mean, maximum value and minimum value of the local steep heterogeneity distance sequence includes: Get the element mean, maximum and minimum value of the local steep heterogeneity distance sequence; Calculate the square of the difference between each element of the local steep heterogeneity distance sequence and the mean of the element, obtain the mean of the square of the difference of all elements of the local steep heterogeneity distance sequence and then take the square root; obtain the difference between the maximum value and the minimum value, and use the difference as the negative exponent of an exponential function with a natural constant as the base; The ratio of the square root to the calculation result of the exponential function is used as the distance sequence disorder index; The step of obtaining the adaptive filtering window size of each region according to the distance sequence disorder index includes: Obtaining a ratio of a distance sequence disorder index to a preset adjustment parameter, obtaining a calculation result of an exponential function with the ratio as an exponent and a natural constant as a base, recording the result as a first result, and calculating a difference between 1 and the first result; Preset a partition coefficient and a basic window coefficient for each region, obtain the product of the partition coefficient and the difference, and use the sum of the basic window coefficient and the product as the adaptive filtering window size of the region.

2. A land remote sensing image correction and enhancement method as claimed in claim 1, characterized in that: The acquisition of land cloud and fog remote sensing images of various regions includes: The grayscale image of the land cloud remote sensing image is obtained, and the canny edge detection algorithm is used to detect the edge of the grayscale image, and a binary image is output. The edge pixels divide the land cloud remote sensing image into various regions.

3. A land remote sensing image correction and enhancement method as claimed in claim 1, characterized in that: The acquisition of land cloud and fog remote sensing images of various regions includes: Obtain a grayscale image of the land cloud remote sensing image, and use the grayscale image as the input of the semantic segmentation network. The output of the semantic segmentation network is a semantic segmentation rendering, where the semantic segmentation rendering is a binary image that divides the land cloud remote sensing image into various regions.

4. A land remote sensing image correction and enhancement method as claimed in claim 2 or 3, characterized in that: The method of constructing the local steepness heterogeneity coefficient of each pixel point in the region according to the difference between the steepness index inside the window of each sliding window in the region and the steepness index inside the region includes: A local window of w*w is constructed with each pixel in the region as the center, where w is the side length of the local window. The internal steepness index of the window in the local window is calculated, and the absolute value of the difference between the internal steepness index of the window and the internal steepness index of the region is obtained. The absolute value of the difference is used as the local steepness heterogeneity coefficient of the central pixel of the local window.

5. A land remote sensing image correction and enhancement method as claimed in claim 4, characterized in that: The method of combining the distance relationship between the elements in the sort pixel set of each region to obtain a highly heterogeneous pixel sequence includes: For the sort pixel set of each region; Randomly select any pixel point in the sort pixel set as a random point, calculate the Manhattan distance between other pixel points in the sort pixel set and the random point, and use the pixel point with the smallest Manhattan distance to the random point as a new random point. Similarly, traverse the sort pixel set to obtain each random point, and compose all random points into a highly heterogeneous pixel sequence.

6. A land remote sensing image correction and enhancement method as claimed in claim 5, characterized in that: The method of obtaining a local steep heterogeneity distance sequence according to the distance between each element in the high heterogeneity pixel sequence includes: The Manhattan distances between adjacent elements in the high heterogeneity pixel sequence are obtained, and the Manhattan distances between all adjacent elements in the high heterogeneity pixel sequence are used as the elements of the local steep heterogeneity distance sequence.

Images