A land use classification method based on image segmentation
By constructing distance measurement adjustment factors in superpixel segmentation algorithm, using feature information of urban remote sensing grayscale images, the image segmentation process is improved, the problem of low accuracy in traditional superpixel segmentation algorithm is solved, and the accuracy of segmentation of land use areas is improved , providing more reliable data support for land use classification.
Patent Information
- Application Number
- CN202411237095.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-05
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2044-09-05
AI Technical Summary
Traditional superpixel segmentation algorithms ignore texture feature information in urban remote sensing images, resulting in low accuracy of image segmentation results, which is prone to oversegmentation, affecting the optimization of land space planning and layout.
By acquiring urban remote sensing grayscale images, determine the characteristic values of similar land use and the characteristic index of the connection domain of each pixel point, combine the gradient amplitude, build a distance measurement adjustment factor, improve the distance measurement method in the superpixel segmentation algorithm, and enhance the accuracy of image segmentation sex.
The accuracy of the segmentation of various land use areas is improved, and the impact of inaccuracies on classification results is avoided, and more reliable land use classification data is provided.
Smart Images

Figure CN118781491B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of image segmentation, and in particular to a land use classification method based on image segmentation. Background Art
[0002] With the acceleration of urbanization, the importance of national land space planning and layout is increasing, and the requirements for optimizing national land space planning and layout are getting higher and higher. At present, with the development of remote sensing image processing technology, image segmentation technology is used to segment urban remote sensing images to obtain various land use areas, and then to achieve reasonable classification of various land use areas. However, due to the inaccuracy of various land use areas obtained by image segmentation technology, the classification results of various land use areas are often affected.
[0003] Superpixel segmentation algorithms are often used to segment remote sensing images. The algorithm has low computational complexity and is generally applicable to large-scale remote sensing images. However, in traditional superpixel segmentation, the distance measurement method only considers color distance and spatial distance, but lacks texture feature information in remote sensing images, which is prone to over-segmentation, resulting in low accuracy of various land use areas obtained by superpixel segmentation. Summary of the invention
[0004] In order to solve the above technical problems, the present application provides a land use classification method based on image segmentation to solve the existing problems.
[0005] The land use classification method based on image segmentation in this application adopts the following technical solution:
[0006] An embodiment of the present application provides a land use classification method based on image segmentation, the method comprising the following steps:
[0007] Obtain urban remote sensing grayscale images;
[0008] Determine the similar land use characteristic value of each pixel point according to the local grayscale difference and grayscale distribution of each pixel point in the urban remote sensing grayscale image; obtain a connected domain extraction window with a preset side length with each pixel point as the center, and determine the similar connected domain characterization index according to the area difference of any two connected domains in the connected domain extraction window of each pixel point combined with the discrimination value of any two connected domains, wherein a connected domain discrimination criterion is constructed based on the relationship between the any two connected domains to obtain the discrimination value between the connected domains; determine the similar land use connected domain characteristic index according to the similar connected domain characterization index and the distribution of similar land use characteristic values of pixel points in the local neighborhood;
[0009] Perform edge detection on the similar land use characteristic values of all pixel points to obtain the gradient amplitude of each pixel point, and determine the category edge distance characteristic value of each pixel point according to the gradient amplitude difference and distribution of each pixel point; determine the distance measurement adjustment factor of each pixel point according to the similar land use characteristic value of each pixel point, the similar land use connected domain characteristic index and the category edge distance characteristic value;
[0010] The distance metric in the image segmentation process is improved based on the distance metric adjustment factor, and the distance metric characteristic parameters between any two pixels are obtained. The urban remote sensing grayscale image is segmented by combining the superpixel segmentation algorithm to obtain various types of land use area images.
[0011] Preferably, the method for determining the characteristic value of the same type of land use is:
[0012] Taking each pixel as the center, a window with a preset side length is obtained, and the grayscale values of all pixels in the window are combined into a grayscale representation sequence of the pixel at the center of the window;
[0013] Determine the land use grayscale variation coefficient of each pixel in the grayscale representation sequence based on the grayscale difference between the pixels in the grayscale representation sequence of each pixel;
[0014] Calculate the information entropy of the grayscale representation sequence, obtain the absolute value of the difference in the land grayscale variation coefficient of any two adjacent pixels in the grayscale representation sequence, calculate the product of the information entropy and the absolute value, and round up the cumulative sum of the products calculated for all any two adjacent pixels in the grayscale representation sequence as the similar land feature value of the corresponding pixel in the grayscale representation sequence.
[0015] Preferably, the land grayscale variation coefficient is the average of the absolute values of the grayscale differences between each pixel in the grayscale characterization sequence and all other pixels in the sequence.
[0016] Preferably, the expression of the characterization index of the same type of connected domain is:
[0017] ;in, is the representation index of the same connected domain of the x-th pixel, Extract the number of connected domains in the window for the connected domain of the x-th pixel. and are the areas of the qth and rth connected domains in the connected domain extraction window of the xth pixel, respectively. is the discriminant value between the qth and rth connected domains, To adjust the parameters.
[0018] Preferably, the calculation formula corresponding to the connected domain judgment criterion is: ;in, is the discriminant value between the qth and rth connected domains.
[0019] Preferably, the method for determining the characteristic index of the connectivity domain of the same type of land is:
[0020] Calculate the normalized value of the coefficient of variation of the similar land use characteristic values of all pixels in the connected domain extraction window of each pixel:
[0021] The sum of the similar connected domain characterization index of each pixel point and a preset number is obtained, the ratio of the normalized value to the sum is calculated, and the difference between the number 1 and the ratio is used as the similar land use connected domain feature index of each pixel point.
[0022] Preferably, the method for determining the category edge distance feature value is:
[0023] Calculate the mean of the absolute values of the differences between the gradient amplitudes of each pixel and all the pixels in its connected domain extraction window, and multiply the mean with the mean square error of the gradient amplitudes of all the pixels in the connected domain extraction window of each pixel as the category edge distance feature value of each pixel.
[0024] Preferably, the distance metric adjustment factor is the normalized result of the product of the characteristic value of the same type of land use, the characteristic index of the connected domain of the same type of land use, and the characteristic value of the category edge distance.
[0025] Preferably, the method for determining the distance measurement characteristic parameter between any two pixel points is:
[0026] The Euclidean distance of the distance metric adjustment factor of the arbitrary two pixel points is calculated, and the distance metric feature parameter between the arbitrary two pixel points is obtained by combining the distance metric between the arbitrary two pixel points in the traditional superpixel segmentation algorithm.
[0027] Preferably, the acquisition of the distance metric characteristic parameter further includes: multiplying the distance metric obtained by the traditional superpixel segmentation algorithm by the Euclidean distance to obtain the distance metric characteristic parameter between any two pixel points.
[0028] This application has at least the following beneficial effects:
[0029] This application is based on the analysis of the grayscale characteristics of similar areas, constructs characteristic values of similar land use, which are used to characterize the characteristics of different types of land use areas, avoid the sub-areas with grayscale changes in similar land use areas, and the problem of incorrect segmentation of existing image segmentation, thereby improving the accuracy of subsequent segmentation; based on the analysis of connected domains in different land use areas, a connected domain characteristic index of similar land use is constructed based on the characteristic values of similar land use, characterizes the connected domain characteristics of different land use areas, and solves the problem that there are many connected domains in similar residential land, resulting in poor segmentation effect for the entire land use area; based on combining the connected domain characteristics of different land use areas and the category edge distance characteristics, a distance metric adjustment factor is constructed to improve the distance metric method in the superpixel segmentation algorithm and improve the accuracy of superpixel segmentation. The present invention improves the distance metric method in the superpixel segmentation algorithm based on the distance metric adjustment factor, improves the accuracy of segmentation of various types of land use areas, thereby avoiding the impact on the classification results of various types of land use areas due to the poor accuracy of segmentation of various types of land use areas. BRIEF DESCRIPTION OF THE DRAWINGS
[0030] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present application 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 application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0031] Figure 1 A flowchart of the steps of a land use classification method based on image segmentation provided for this application;
[0032] Figure 2 Schematic diagram of the process of improving the distance metric for superpixel segmentation algorithm. DETAILED DESCRIPTION
[0033] In order to further explain the technical means and effects adopted by the present application to achieve the predetermined invention purpose, the following is a detailed description of the specific implementation method, structure, features and effects of a land use classification method based on image segmentation proposed in the present application in combination 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.
[0034] 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 application belongs.
[0035] The following is a detailed description of a specific scheme of a land use classification method based on image segmentation provided by the present application in conjunction with the accompanying drawings.
[0036] An embodiment of the present application provides a land use classification method based on image segmentation. Specifically, the following land use classification method based on image segmentation is provided. Figure 1 , the method comprises the following steps:
[0037] Step S001: collect town remote sensing images, and pre-process the town remote sensing images to obtain town remote sensing grayscale images.
[0038] In order to avoid affecting the results of optimizing the configuration of various land use areas due to inaccurate segmentation of various land use areas, it is necessary to improve the accuracy of image segmentation to obtain various land use areas.
[0039] This embodiment acquires a remote sensing image of a town by collecting remote sensing images of the town. In order to facilitate the subsequent analysis of the segmentation of various land use areas and improve the accuracy of image segmentation, the town remote sensing image is preprocessed, the town remote sensing image is used as the input of the non-local mean filtering algorithm, the output of the non-local mean filtering algorithm is used as the town remote sensing pure image, the town remote sensing pure image is used as the input of the grayscale average algorithm, and the output of the grayscale average algorithm is used as the town remote sensing grayscale image, wherein the non-local mean filtering algorithm and the grayscale average algorithm are well-known technologies, and the specific process is not repeated.
[0040] So far, the preprocessed urban remote sensing grayscale image.
[0041] Step S002, constructing the characteristic value of similar land use to obtain the town category characteristic image, obtaining the characteristic index of the connected domain of similar land use according to the town category characteristic image, calculating the gradient amplitude, constructing the category edge distance characteristic value, and obtaining the distance measurement adjustment factor according to the characteristic index of the connected domain of similar land use and the category edge distance characteristic value.
[0042] The principle of superpixel segmentation algorithm is an algorithm based on similarity segmentation. However, the distance measurement method in the traditional superpixel segmentation algorithm is only based on color distance and spatial distance, ignoring the rich texture feature information in remote sensing images, which often leads to low accuracy of image segmentation for various land use areas. Therefore, in order to improve the accuracy of segmentation of various land use areas, it is necessary to improve the distance measurement method in the traditional superpixel segmentation algorithm.
[0043] Generally, different types of land use areas in urban remote sensing grayscale images have different grayscale features, while the same type of land use areas have similar grayscale representation features. Therefore, in order to more clearly obtain the grayscale representation features of different types of land use areas, it is necessary to extract grayscale representation features based on urban remote sensing grayscale images.
[0044] Specifically, taking each pixel in the urban remote sensing grayscale image as the center, set The size window is used as the grayscale representation window of each pixel, and the implementer can determine the size value according to the actual situation. In order to more accurately reflect the grayscale changes of different types of land use areas, for the grayscale representation window of each pixel, the sequence composed of the grayscale values of all pixels in the grayscale representation window is used as the grayscale representation sequence of each pixel.
[0045] Based on the above analysis, calculate the characteristic value of similar land use for each pixel : ; ;in, is the land grayscale change coefficient of the jth element in the grayscale representation sequence of the xth pixel, is the ceiling function, is the number of elements in the grayscale representation sequence of the x-th pixel, and are the jth and gth element values in the grayscale representation sequence of the xth pixel, respectively; is the characteristic value of the same type of land use at the x-th pixel point, is the information entropy of the land grayscale change coefficient of all elements in the grayscale representation sequence of the x-th pixel, and They are the land use grayscale change coefficients of the jth and j-1th elements in the grayscale representation sequence of the xth pixel point respectively.
[0046] It should be understood that since different types of land use areas have different grayscale changes, in order to distinguish more accurately, the grayscale value changes The larger the value is, the greater the grayscale change characteristic is in the land use area where the pixel is located, and the greater the land use grayscale change coefficient is; and the information entropy of the land use grayscale change coefficient is The larger the value is, the richer the information of regional grayscale change is, and the grayscale change coefficient of land use is different. The larger it is, the clearer the grayscale change characteristics of the land use area are, and the better it can characterize the characteristics of similar land use areas, the larger the characteristic value of similar land use areas is.
[0047] Furthermore, based on the above analysis, the characteristic value of similar land use for each pixel is calculated to avoid the problem of sub-regions with grayscale changes in similar land use areas and incorrect segmentation of existing image segmentation. At this time, the characteristic value of similar land use can characterize the pixel characteristics in each type of land use area.
[0048] In addition, land use areas usually have clearer land use characteristics, while urban remote sensing grayscale images often reflect poor land use conditions. In order to make the image reflect different land use characteristics more clearly, the grayscale value of each pixel in the urban remote sensing grayscale image is replaced with the characteristic value of the same type of land use for each pixel. All pixels in the urban remote sensing grayscale image are traversed and replaced one by one. The result of the replacement of the urban remote sensing grayscale image is used as the urban category feature image.
[0049] Furthermore, under normal circumstances, the town category feature image has relatively clear connected domain features, but some areas of the same type of land use often do not have a connected domain. For example, residential land often makes the connected domain more complex due to changes in the color of the roofs of houses of different residents and changes in the layout of houses, and it is impossible to clearly reflect the residential land.
[0050] Land use areas often form a larger connected domain, while residential land is often composed of more small connected domains, and the distances between the connected domains are relatively close.
[0051] In order to improve the accuracy of subsequent segmentation, each pixel in the town category feature image is set as the center. The window is used as the connected domain extraction window for each pixel, the matrix composed of the same type of land use eigenvalues of all pixels in the connected domain extraction window of each pixel is used as the category eigenvalue matrix of each pixel, and the category eigenvalue matrix of each pixel is used as the input of the grayscale region size matrix algorithm. The grayscale region size matrix algorithm outputs the number of connected domains in the connected domain extraction window of each pixel and the area of each connected domain. The grayscale region size matrix algorithm is a well-known technology and the specific process will not be repeated here.
[0052] Based on the above analysis, the characteristic index of the connected domain of similar land use at each pixel is calculated. : ; ;in, is the representation index of the same connected domain of the x-th pixel, Extract the number of connected domains in the window for the connected domain of the x-th pixel. and are the areas of the qth and rth connected domains in the connected domain extraction window of the xth pixel, respectively. As a criterion, When , the discriminant value corresponding to the qth and rth connected domains is 0; when When , the corresponding discriminant value is 1. To adjust the parameters and avoid the denominator being 0, remove the empirical value 1; is the characteristic index of the connected domain of the same type of land use at the x-th pixel point, is the normalization function, The coefficient of variation of the characteristic values of similar land use of all pixels in the connected domain extraction window of the x-th pixel is obtained.
[0053] It can be understood that the smaller the difference between the sizes of connected domains within the connected domain extraction window, the more it conforms to the connected domain characteristics of similar land use areas, and the larger the size of a single connected domain, the more likely it is a connected domain of similar land use, and the larger the similar connected domain representation index; the larger the similar connected domain representation index and the larger the coefficient of variation, the greater the possibility that similar land use areas are divided, and at this time, the more connected domain characteristics it has similar land use areas, and the larger the similar land use connected domain characteristic index.
[0054] Furthermore, based on the above analysis, the feature index of the connected domain of the same type of land use for each pixel is obtained. The distance measurement method in the traditional superpixel segmentation algorithm is only based on color distance and spatial distance, resulting in poor segmentation accuracy. In order to improve the accuracy of segmentation, it is necessary to extract the edge features of various land use areas.
[0055] Specifically, the town category feature image is used as the input of the Sobel edge detection algorithm, and the output of the Sobel edge detection algorithm is used as the gradient amplitude of each pixel in the town category feature image. The Sobel edge detection algorithm is a well-known technology and the specific process is not repeated here.
[0056] Extract the window based on the connected domain of each pixel and calculate the category edge distance feature value of each pixel : ;in, is the category edge distance feature value of the x-th pixel, Extract the gradient amplitude of the k-th pixel in the window for the connected domain of the x-th pixel, is the gradient amplitude of the x-th pixel, K is the number of all neighboring pixels in the connected domain extraction window of the x-th pixel, Represents the mean square error of the gradient amplitudes of all pixels in the connected domain extraction window of the x-th pixel.
[0057] It should be understood that since the category differences in the location of the land division area are often large, there are large differences in the gradient amplitudes of the category edge pixels at this time, and the larger the mean square error, the more likely it is to be at the edge of the land area, and the larger the category edge distance eigenvalue.
[0058] Furthermore, based on the characteristic index of the connected domain of the same type of land and the characteristic value of the class edge distance calculated above, the distance metric adjustment factor of each pixel is calculated to improve the distance metric method in the traditional superpixel segmentation algorithm. The calculation formula is: ; In the formula, is the distance metric adjustment factor for the x-th pixel, is the normalization function, is the category edge distance feature value of the x-th pixel, is the characteristic index of the connected domain of the same type of land use at the x-th pixel point, is the characteristic value of the same type of land use at the x-th pixel.
[0059] The normalized value of the product of the representation index of the same type of connected domain and the characteristic value of the category edge distance is used as the weight for calculating the distance metric adjustment factor. The larger the weight value, the greater the role it plays in image segmentation. In addition, the same type of land use characteristics reflect the size of the distance metric adjustment factor, making the improved distance metric more accurate.
[0060] At this point, the distance metric adjustment factor for each pixel is obtained.
[0061] Step S003, obtaining distance metric feature parameters according to the distance metric adjustment factor, and obtaining various types of land use area images based on the distance metric feature parameters using a superpixel segmentation algorithm.
[0062] Furthermore, the distance metric in the superpixel segmentation algorithm is improved based on the distance metric adjustment factor, and the distance metric characteristic parameters between any two pixels after the improvement are calculated: ;in, is the distance measurement feature parameter between the vth and hth pixels, is the distance between the vth and hth pixels in the traditional superpixel segmentation algorithm. It should be noted that the distance between two pixels in the traditional superpixel segmentation algorithm may be the Euclidean distance between the two pixels. The implementer may also use other existing distance measurement methods. is the Euclidean distance function, and are the distance metric adjustment factors for the vth and hth pixels respectively.
[0063] Furthermore, in this embodiment, the urban remote sensing grayscale image is used as the input of the superpixel segmentation algorithm, the calculation of the distance measurement feature parameter is used as the distance measurement method in the superpixel segmentation algorithm, and the output of the superpixel segmentation algorithm is used as various land use area images, wherein the superpixel segmentation algorithm is a well-known technology, and the specific process is not repeated here. The schematic diagram of the improved process of the superpixel segmentation algorithm distance measurement is as follows: Figure 2 shown.
[0064] At this point, according to the above process of this embodiment, accurate segmentation of urban remote sensing grayscale images can be achieved, and land use can be classified based on the segmentation results, providing reliable data support for land use distribution optimization and completing the accurate division of land areas.
[0065] It should be noted that the above sequence of the embodiments of the present application is for description only 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.
[0066] 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.
[0067] 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 use classification method based on image segmentation, It is characterized in that The method comprises the following steps: Obtain urban remote sensing grayscale images; Determine the similar land use characteristic value of each pixel point according to the local grayscale difference and grayscale distribution of each pixel point in the urban remote sensing grayscale image; obtain a connected domain extraction window with a preset side length with each pixel point as the center, and determine the similar connected domain characterization index according to the area difference of any two connected domains in the connected domain extraction window of each pixel point combined with the discrimination value of any two connected domains, wherein a connected domain discrimination criterion is constructed based on the relationship between the any two connected domains to obtain the discrimination value between the connected domains; determine the similar land use connected domain characteristic index according to the similar connected domain characterization index and the distribution of similar land use characteristic values of pixel points in the local neighborhood; Perform edge detection on the similar land use characteristic values of all pixel points to obtain the gradient amplitude of each pixel point, and determine the category edge distance characteristic value of each pixel point according to the difference and distribution of the gradient amplitude of each pixel point; determine the distance measurement adjustment factor of each pixel point according to the similar land use characteristic value of each pixel point, the similar land use connected domain characteristic index and the category edge distance characteristic value; The distance metric in the image segmentation process is improved based on the distance metric adjustment factor, and the distance metric characteristic parameters between any two pixels are obtained. The urban remote sensing grayscale image is segmented by combining the superpixel segmentation algorithm to obtain various types of land use area images.
2. A land use classification method based on image segmentation as claimed in claim 1, It is characterized in that The method for determining the characteristic value of the same type of land use is as follows: Taking each pixel as the center, a window with a preset side length is obtained, and the grayscale values of all pixels in the window are combined into a grayscale representation sequence of the pixel at the center of the window; Determine the land use grayscale variation coefficient of each pixel in the grayscale representation sequence based on the grayscale difference between the pixels in the grayscale representation sequence of each pixel; Calculate the information entropy of the grayscale representation sequence, obtain the absolute value of the difference in the land grayscale variation coefficient of any two adjacent pixels in the grayscale representation sequence, calculate the product of the information entropy and the absolute value, and round up the cumulative sum of the products calculated for all any two adjacent pixels in the grayscale representation sequence as the similar land feature value of the corresponding pixel in the grayscale representation sequence.
3. A land use classification method based on image segmentation as claimed in claim 2, It is characterized in that The land grayscale variation coefficient is the average of the absolute values of the grayscale differences between each pixel in the grayscale characterization sequence and all other pixels in the sequence.
4. The land use classification method based on image segmentation as claimed in claim 1, It is characterized in that The expression of the characterization index of the same type of connected domain is: ;in, is the representation index of the same connected domain of the x-th pixel, Extract the number of connected domains in the window for the connected domain of the x-th pixel. and are the areas of the qth and rth connected domains in the connected domain extraction window of the xth pixel, respectively. is the discriminant value between the qth and rth connected domains, To adjust the parameters.
5. The land use classification method based on image segmentation as claimed in claim 1, It is characterized in that The calculation formula corresponding to the connected domain judgment criterion is: ;in, is the discriminant value between the qth and rth connected domains.
6. A land use classification method based on image segmentation as claimed in claim 1, It is characterized in that The method for determining the characteristic index of the connectivity domain of the same type of land is as follows: Calculate the normalized value of the coefficient of variation of the similar land use characteristic values of all pixels in the connected domain extraction window of each pixel: The sum of the similar connected domain characterization index of each pixel point and a preset number is obtained, the ratio of the normalized value to the sum is calculated, and the difference between the number 1 and the ratio is used as the similar land use connected domain feature index of each pixel point.
7. The land use classification method based on image segmentation as claimed in claim 1, It is characterized in that The method for determining the category edge distance feature value is: Calculate the mean of the absolute values of the differences between the gradient amplitudes of each pixel and all the pixels in its connected domain extraction window, and multiply the mean with the mean square error of the gradient amplitudes of all the pixels in the connected domain extraction window of each pixel as the category edge distance feature value of each pixel.
8. The land use classification method based on image segmentation as claimed in claim 1, It is characterized in that The distance metric adjustment factor is the normalized result of the product of the characteristic value of the same type of land use, the characteristic index of the connected domain of the same type of land use, and the characteristic value of the category edge distance.
9. The land use classification method based on image segmentation as claimed in claim 1, It is characterized in that The method for determining the distance measurement characteristic parameter between any two pixel points is as follows: The Euclidean distance of the distance metric adjustment factor of the arbitrary two pixel points is calculated, and the distance metric feature parameter between the arbitrary two pixel points is obtained by combining the distance metric between the arbitrary two pixel points in the traditional superpixel segmentation algorithm.
10. A land use classification method based on image segmentation as claimed in claim 9, It is characterized in that The acquisition of the distance metric characteristic parameter further includes: multiplying the distance metric obtained by the traditional superpixel segmentation algorithm by the Euclidean distance to obtain the distance metric characteristic parameter between any two pixel points.
