Remote sensing image segmentation method, device and equipment and computer readable storage medium

Abstract

The application provides a remote sensing image segmentation method, device and equipment and a computer readable storage medium, and belongs to the technical field of land surveying and mapping. The method comprises the following steps: acquiring remote sensing images and point cloud data of a target area; performing fusion processing on the remote sensing images and the point cloud data to obtain fused images, wherein the fused images comprise the elevation information of each pixel point; and performing segmentation on the fused images based on the pixel information of each pixel point in the fused images to obtain a segmentation result, wherein the pixel information comprises the elevation information and color information of the pixel point. Through the above method, the elevation information of a certain region can be assigned to the remote sensing images of the corresponding region, the segmentation of the remote sensing images is further realized, and finally, the super-pixel image is obtained to facilitate subsequent application. The method improves the problem in the prior art that many algorithms need to be used for segmentation, the coding requirement of technical personnel is high, and the image processing complexity is too high.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of land surveying and mapping, in particular to a remote sensing image segmentation method and device, equipment and a computer readable storage medium. BACKGROUND

[0002] In the field of land surveying and mapping, remote sensing images need to be collected and segmented to complete various surveying and mapping work based on the segmented remote sensing images.

[0003] In the prior art, the method of machine learning is generally used to realize the segmentation of remote sensing images. Specifically, an automatic interpretation task of remote sensing images based on machine learning is established, and various semantic segmentations are performed after the target is identified to realize the segmentation of remote sensing images.

[0004] However, the method of the prior art involves digital image processing, spatial domain image enhancement processing algorithms, and also requires the use of neural network algorithms, random forest machine learning algorithms and the like. Therefore, the method of the prior art has the problem of high complexity. SUMMARY

[0005] The purpose of the present application is to provide a remote sensing image segmentation method, device, equipment and computer readable storage medium, which can improve the remote sensing image segmentation effect and reduce the complexity of remote sensing image segmentation processing.

[0006] Embodiments of the present application are implemented as follows:

[0007] In a first aspect, the present application provides a remote sensing image segmentation method, comprising:

[0008] obtaining remote sensing images and point cloud data of a target area;

[0009] performing fusion processing on the remote sensing images and the point cloud data to obtain a fused image, the fused image including elevation information of each pixel point;

[0010] segmenting the fused image based on pixel information of each pixel point in the fused image to obtain a segmentation result, the pixel information including elevation information and color information of the pixel point.

[0011] Optionally, the fusion processing of the remote sensing images and the point cloud data to obtain a fused image comprises:

[0012] converting the point cloud data into digital surface model data;

[0013] performing geometric correction on the remote sensing images based on the digital surface model data to obtain a corrected image;

[0014] fusing the digital surface model data and the corrected image to obtain the fused image.

[0015] Optionally, the converting the point cloud data into the digital surface model data comprises:

[0016] performing coordinate conversion on the point cloud data to obtain converted point cloud data;

[0017] performing classification on the converted point cloud data to obtain a plurality of classified point cloud sets, each of the point cloud sets comprising point cloud data of one type;

[0018] converting the plurality of classified point cloud sets into the digital surface model data.

[0019] Optionally, the fusing the digital surface model data and the corrected image to obtain the fused image comprises:

[0020] performing image registration on the digital surface model data and the corrected image;

[0021] performing data association processing on the registered digital surface model data and the corrected image to obtain a corresponding relationship between pixel points of the registered digital surface model data and the corrected image;

[0022] assigning, according to the corresponding relationship between pixel points of the registered digital surface model data and the corrected image, elevation information in the registered digital surface model data to each pixel point in the corrected image, and taking the corrected image with the assigned elevation information as the fused image.

[0023] Optionally, the segmenting the fused image based on pixel information of each pixel point in the fused image to obtain a segmentation result comprises:

[0024] segmenting the fused image based on pixel information of each pixel point in the fused image using a superpixel segmentation algorithm to obtain a segmentation result.

[0025] Optionally, the segmenting the fused image based on pixel information of each pixel point in the fused image using a superpixel segmentation algorithm to obtain a segmentation result comprises:

[0026] determining a plurality of initial seed centers in the fused image, each of the initial seed centers being a pixel point in the fused image;

[0027] According to pixel information of each pixel point in the fused image, pixel distances of each associated pixel point in a neighborhood corresponding to each initial seed center are determined, and each initial seed center is adjusted according to the pixel distances of each associated pixel point in the corresponding neighborhood, to obtain at least one new seed center. The step is repeatedly executed until the pixel distances of each associated pixel point in a neighborhood corresponding to each new seed center after current adjustment to the initial seed center satisfy a preset condition, and the neighborhood corresponding to each new seed center after current adjustment is taken as a segmentation region.

[0028] Optionally, the determining of the pixel distances of each associated pixel point in the neighborhood corresponding to each initial seed center according to the pixel information of each pixel point in the fused image comprises:

[0029] According to the elevation information of the initial seed center and the elevation information of each associated pixel point associated with the position of the initial seed center, a spatial distance between the initial seed center and each associated pixel point is determined.

[0030] According to the color information of the initial seed center and the color information of each associated pixel point associated with the position of the initial seed center, a color distance between the initial seed center and each associated pixel point is determined.

[0031] According to the spatial distance and the color distance, a pixel distance between the initial seed center and each associated pixel point is determined.

[0032] In a second aspect, the embodiment of the application provides a remote sensing image segmentation device, which is characterized in that the remote sensing image segmentation device comprises:

[0033] An acquisition module is configured to acquire remote sensing image and point cloud data of a target region.

[0034] A fusion module is configured to perform fusion processing on the remote sensing image and the point cloud data to obtain a fused image, wherein the fused image comprises elevation information of each pixel point.

[0035] A segmentation module is configured to perform segmentation on the fused image based on pixel information of each pixel point in the fused image to obtain a segmentation result, wherein the pixel information comprises the elevation information and color information of the pixel point.

[0036] In a third aspect, the embodiment of the application provides an electronic device, which comprises a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, the remote sensing image segmentation method in the first aspect is implemented.

[0037] In a fourth aspect, the present application provides a computer readable storage medium storing a computer program, which, when executed by a processor, implements the remote sensing image segmentation method of the first aspect.

[0038] The beneficial effects of the embodiments of the present application include:

[0039] The remote sensing image segmentation method provided by the embodiments of the present application obtains remote sensing images and point cloud data of a target region, performs fusion processing on the remote sensing images and the point cloud data to obtain fused images, segments the fused images based on pixel information of each pixel point in the fused images, and obtains segmentation results.

[0040] The remote sensing images can also be used to record color information of each ground object, and the point cloud data can be used to indicate the outer surface shape, geometric position information and / or reflection intensity of each ground object. Therefore, by performing fusion processing on the remote sensing images and the point cloud data, the elevation information and color information of each pixel point in the obtained fused images can correspond to indicate the elevation information and color information of each ground object or a part of each ground object, which facilitates subsequent segmentation of the fused images based on the elevation information and color information of each pixel point in the fused images.

[0041] Based on the pixel information of each pixel point in the fused images, the fused images are segmented to obtain segmentation results. Based on the elevation information of each pixel point in the fused images, the height of each ground object and the ups and downs of the terrain in the target region can be determined. Based on the color information of each pixel point in the fused images, the color, brightness, texture and other parameters of each ground object in the target region can be determined. Therefore, based on the elevation information and color information of the fused images, each pixel point with the same or similar elevation information and the same or similar color information in the fused images can be divided into a segmentation region, so that the pixel points with similar or identical pixel information in the fused images can be divided into a segmentation region, thereby achieving segmentation of the fused images.

[0042] In addition, the remote sensing image segmentation method provided by the embodiments of the present application does not need to use a neural network, a random forest or other learning algorithm to identify each ground object in the remote sensing images or perform image enhancement processing in the spatial domain, so that each pixel point with the same or similar elevation information and the same or similar color information in the fused images can be accurately and reliably divided into a segmentation region, thereby achieving the purpose of accurately and reliably segmenting the fused images. In this way, the effect of improving the remote sensing image segmentation effect and reducing the complexity of the remote sensing image segmentation processing can be achieved. BRIEF DESCRIPTION OF DRAWINGS

[0043] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some of the embodiments of the present application, and therefore should not be regarded as limiting the scope, and for those skilled in the art, other related drawings can also be obtained without creative labor.

[0044] Figure 1 The flow chart of the first remote sensing image segmentation method provided by the embodiments of the present application;

[0045] Figure 2 The flow chart of the second remote sensing image segmentation method provided by the embodiments of the present application;

[0046] Figure 3 The flow chart of the third remote sensing image segmentation method provided by the embodiments of the present application;

[0047] Figure 4 The flow chart of the fourth remote sensing image segmentation method provided by the embodiments of the present application;

[0048] Figure 5 The flow chart of the fifth remote sensing image segmentation method provided by the embodiments of the present application;

[0049] Figure 6 The flow chart of the sixth remote sensing image segmentation method provided by the embodiments of the present application;

[0050] Figure 7 The structural schematic diagram of a remote sensing image segmentation device provided by the embodiments of the present application;

[0051] Figure 8 The structural schematic diagram of an electronic device provided by the embodiments of the present application. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solutions and advantages of the embodiments of the present application more clear, the following will combine the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are some of the embodiments of the present application, but not all the embodiments. The components of the embodiments of the present application described and shown in the drawings can be arranged and designed in various different configurations.

[0053] Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0054] In the prior art, a method of machine learning is commonly used to implement the segmentation of remote sensing images. Specifically, an automatic interpretation task of remote sensing images based on machine learning is established, and the segmentation of remote sensing images is implemented by identifying targets first and then performing various semantic segmentation. However, the method of the prior art involves algorithms such as digital image processing and spatial domain image enhancement processing, and also needs to use machine learning algorithms such as neural network algorithms and random forests. Therefore, the method of the prior art has the problem of high complexity.

[0055] Therefore, the embodiments of the present application provide a remote sensing image segmentation method, which obtains remote sensing images and point cloud data of a target area, performs fusion processing on the remote sensing images and the point cloud data to obtain a fused image, performs segmentation on the fused image based on pixel information of each pixel point in the fused image, and obtains a segmentation result, thereby achieving the effects of improving the remote sensing image segmentation effect and reducing the complexity of remote sensing image segmentation processing.

[0056] The embodiments of the present application take a remote sensing image segmentation method applied in land surveying and mapping as an example for illustration. However, it is not indicated that the embodiments of the present application can only be applied to remote sensing image segmentation in land surveying and mapping.

[0057] The remote sensing image segmentation method provided by the embodiments of the present application is explained and described in detail below.

[0058] Figure 1 A flowchart of a remote sensing image segmentation method provided by the present application is shown in FIG. 1. The method can be applied to an electronic device, which can be any terminal device or server having a processing function. Referring to FIG. 1, the embodiments of the present application provide a remote sensing image segmentation method, which includes the following steps. Figure 1 The embodiments of the present application provide a remote sensing image segmentation method, which includes the following steps.

[0059] Step 1001: Obtain remote sensing images and point cloud data of a target area.

[0060] Optionally, the target area can be any area selected by a related technical person for land surveying, or can be an area from which a radar can collect remote sensing images and / or point cloud data, and the embodiments of the present application do not limit this.

[0061] Optionally, the remote sensing images can be images obtained by detecting electromagnetic wave information reflected, radiated or scattered by objects by a satellite, a remote sensor or any other possible sensor, that is, the remote sensing images can be photographs recording the sizes of electromagnetic waves of various ground objects, and specifically, the remote sensing images can also be used to record color information of various ground objects. For example, the remote sensing images can be aerial photographs or satellite photographs, and the embodiments of the present application do not limit this.

[0062] Exemplarily, the remote sensing image can be an image in RGB mode, an image in CMYK mode, or an image in Lab mode, and the embodiments of the present application do not make any limitation in this aspect.

[0063] Optionally, the point cloud data can refer to a set of vectors in a three-dimensional coordinate system. The vectors are usually represented in the form of (X, Y, Z) three-dimensional coordinates, and can generally be used to represent the outer surface shape of any object and / or the geometric position information of an object.

[0064] In addition, the point cloud data can be in the form of a data set or in the form of a three-dimensional image, and the embodiments of the present application do not make any limitation in this aspect.

[0065] Exemplarily, the point cloud data can be laser radar point cloud data collected by an airborne laser radar (Light Detection And Ranging, LiDAR for short), or can be various types of point cloud data collected by any other possible means, and the embodiments of the present application do not make any limitation in this aspect.

[0066] For example, the dense high-precision 3D point coordinates of the ground surface can be directly and quickly obtained by emitting and receiving laser pulses by the LiDAR to obtain the point cloud data. In addition, the precision of the point cloud data can be adjusted according to actual needs. Generally, the precision of the point cloud data can be set to 35 3D points per square meter, that is, the point cloud data can indicate 35 3D point coordinates per square meter, or the point cloud data can be set to any other possible precision, and the embodiments of the present application do not make any limitation in this aspect.

[0067] If the point cloud data is laser radar point cloud data obtained by the LiDAR, the point cloud data can also be used to indicate the reflection intensity of each ground object in the target area.

[0068] It is worth noting that the remote sensing image can also specifically refer to an image obtained after correcting a photograph collected by a satellite, a remote sensor, or other sensors. In this way, the errors caused by the imaging rules of remote sensing technology can be eliminated, and the accuracy and reliability of the remote sensing image segmentation method can be improved.

[0069] In this way, the remote sensing image recording the electromagnetic wave size of each ground object in the target area and the point cloud data recording the outer surface shape and / or geometric position information of each ground object in the target area can be obtained, which facilitates subsequent operations.

[0070] Step 1002: performing fusion processing on the remote sensing image and the point cloud data to obtain a fused image.

[0071] Optionally, the fused image includes the elevation information of each pixel.

[0072] Each pixel point can refer to any pixel point in the fused image, or can refer to all pixel points in the fused image.

[0073] Optionally, the elevation information can be calculated from three-dimensional coordinates in the point cloud data indicating the outer surface shape and / or geometric position information of each ground object, or can be obtained from the point cloud data in any other possible manner, which is not limited in the embodiments of the present application.

[0074] The elevation information can refer to the distance of any point on an object represented by the pixel point to an absolute base surface along the vertical direction. The absolute base surface can be the ground, or any reference surface. The embodiments of the present application do not make any limitation in this regard.

[0075] Specifically, the elevation information can be used to indicate the height of each ground object in the target area, and can be used to indicate the ups and downs of the terrain in the target area, which is not limited in the embodiments of the present application.

[0076] Exemplarily, the fusion processing of the remote sensing image and the point cloud data can be performed in the intensity-hue-saturation PAC method, the Pansharp algorithm, the Brovey transform method and the wavelet transform method, etc.

[0077] It is worth noting that since the fused image is obtained by fusing the remote sensing image and the point cloud data, the fused image records each ground object, the electromagnetic wave size of each ground object and the corresponding point cloud data recorded in the remote sensing image, and each pixel point in the fused image can correspond to each ground object or a part of each ground object. Therefore, all pixel points in the fused image can indicate all ground objects and the elevation information of all ground objects in the target area.

[0078] By fusing the remote sensing image and the point cloud data to obtain the fused image, the complementary information between the remote sensing image and the point cloud data can be mined by fusion processing, the low data quality defect of the remote sensing image or the point cloud data can be eliminated, and the data accuracy of the fused image can be improved.

[0079] Step 1003: based on the pixel information of each pixel point in the fused image, the fused image is segmented to obtain a segmentation result.

[0080] Optionally, the pixel information includes the elevation information and color information of each pixel point.

[0081] Optionally, the color information can be used to indicate the texture, color, brightness and other parameters of each pixel point.

[0082] The segmentation result can include a plurality of segmentation regions, and each segmentation region can refer to a part of the fused image after the fused image is segmented. That is, the fused image can be obtained again after each segmented image is combined.

[0083] In addition, the pixel information of each pixel point in the segmentation region is the same or similar, or the correlation degree of the pixel information of each pixel point in the segmentation region is high. The correlation degree can be calculated according to the position of each pixel point in the fused image, the elevation information of each pixel point, and the color information, which is not limited by the embodiments of the present application.

[0084] Notably, since the fused image is obtained by fusing the remote sensing image of the target region and the point cloud data, and the remote sensing image is a photo for recording the size of electromagnetic waves of various ground objects, the remote sensing image can record the color, brightness, and / or texture parameters of each ground object in the target region. Therefore, the fused image can also record the color, brightness, and / or texture parameters of each ground object in the target region, that is, the color information of each pixel point.

[0085] It is worth noting that the height of each ground object in the target region and the ups and downs of the terrain can be determined based on the elevation information of each pixel point in the fused image, and the color, brightness, and texture parameters of each ground object in the target region can be determined based on the color information of each pixel point in the fused image. Therefore, the fused image can be segmented based on the elevation information and color information of the fused image, so that each pixel point with the same or similar elevation information and the same or similar color information in the fused image can be divided into a segmentation region. In this way, the pixel points with similar or identical pixel information in the fused image can be divided into a segmentation region, so as to achieve the segmentation of the fused image.

[0086] Since the point cloud data can be obtained by scanning each ground object in the target region by LiDAR or other devices, and the remote sensing image is originally a photo or image, that is, the remote sensing image originally includes the color, brightness, and texture parameters of each ground object in the target region. Therefore, the fused image can be segmented based on the elevation information and color information of the fused image, so that a large amount of identification work and image enhancement processing of the remote sensing image can be avoided, and each pixel point with the same or similar elevation information and the same or similar color information in the fused image can be accurately and reliably divided into a segmentation region, so as to achieve the purpose of accurately and reliably segmenting the fused image.

[0087] In the embodiment of the present application, the remote sensing image and the point cloud data of the target area are acquired, the remote sensing image and the point cloud data are fused to obtain a fused image, the fused image is segmented based on the pixel information of each pixel in the fused image to obtain a segmentation result.

[0088] The remote sensing image can also be used to record the color information of each ground object, and the point cloud data can be used to indicate the outer surface shape, geometric position information and / or the reflection intensity of each ground object. Therefore, the height information and the color information of each ground object or part of each ground object in the fused image obtained by fusing the remote sensing image and the point cloud data can be indicated, which facilitates subsequent segmentation of the fused image based on the height information and the color information of each pixel in the fused image.

[0089] The fused image is segmented based on the pixel information of each pixel in the fused image to obtain a segmentation result. The height of each ground object and the ups and downs of the terrain in the target area can be determined based on the height information of each pixel in the fused image. The color, brightness, texture and other parameters of each ground object in the target area can be determined based on the color information of each pixel in the fused image. Therefore, the fused image can be segmented based on the height information and the color information of the fused image, so that each pixel in the fused image with the same or similar height information and the same or similar color information can be divided into a segmentation region. In this way, the pixels with similar or identical pixel information in the fused image can be divided into a segmentation region, so as to achieve segmentation of the fused image.

[0090] In addition, the remote sensing image segmentation method provided by the embodiment of the present application does not need to identify each ground object in the remote sensing image or perform image enhancement processing in the spatial domain through a neural network, a random forest or other learning algorithm. Therefore, each pixel in the fused image with the same or similar height information and the same or similar color information can be accurately and reliably divided into a segmentation region, so as to accurately and reliably segment the fused image. In this way, the effect of remote sensing image segmentation can be improved, and the complexity of remote sensing image segmentation processing can be reduced.

[0091] In a possible implementation manner, referring to Figure 2 The remote sensing image and the point cloud data are fused to obtain a fused image, including:

[0092] Step 1004: converting the point cloud data into digital surface model data.

[0093] Optionally, the digital surface model (DSM) can refer to a digital model used for geometric correction of the remote sensing image. For example, the digital surface model can be a quadratic polynomial model, an affine transformation model, a local triangulation network model, or any other model that can be used for geometric correction.

[0094] The digital surface model data can refer to data converted from the point cloud data to facilitate geometric correction of the remote sensing image by the digital surface model. The digital surface model data can also refer to data obtained by classifying the point cloud data. Embodiments of the present application do not limit this.

[0095] In addition, the point cloud data can be classified by corresponding classification software or classification programs. For example, the point cloud data can be classified by point cloud data processing software LiDAR-DP, or the point cloud data can be classified by a related technical person. Embodiments of the present application do not limit this.

[0096] Since the point cloud data can be in the form of a data set or in the form of a three-dimensional image, the digital surface model data obtained by converting the point cloud data can be in the form of a data set or in the form of a three-dimensional image. Embodiments of the present application do not limit this.

[0097] In this way, the remote sensing image can be conveniently geometrically corrected.

[0098] Step 1005: Geometrically correcting the remote sensing image based on the digital surface model data to obtain a corrected image.

[0099] Optionally, the corrected image and the target region contained in the remote sensing image can be the same. The electromagnetic wave size of each ground object, the color information of each ground object, and the number of each pixel in the corrected image and the remote sensing image can also be the same. Each pixel in the corrected image and each pixel in the remote sensing image can correspond to the same ground object or different ground objects. Embodiments of the present application do not limit this.

[0100] In addition, the geometric correction operation can refer to an operation of correcting geometric deformation in the remote sensing image due to factors such as earth curvature, changes in terrain, and air refraction using ground control points and the digital surface model, and projecting the geometrically corrected remote sensing image onto a plane. The ground control point can be a point used as a reference point for geometric correction in the target region.

[0101] Exemplarily, the operation of geometric correction can specifically include an operation of adjusting the remote sensing image from central projection to orthographic projection, and an operation of adjusting the height difference projection for the terrain undulations or height differences of the objects in the target area.

[0102] For example, one or more ground control points can be selected from the remote sensing image or the vector data corresponding to the remote sensing image, or determined according to the measurement results of the relevant technical personnel in the target area, and the embodiments of the present application do not limit this.

[0103] Then, the positions and information of the pixel points corresponding to the ground control points in the remote sensing image are determined, the point cloud data or elevation information corresponding to the ground control points is read from the digital surface model data, and the positions and information of the pixel points corresponding to the ground control points in the remote sensing image are geometrically matched with the point cloud data or elevation information corresponding to the ground control points, and the remote sensing image is corrected according to the matching result.

[0104] For example, the ENVI5.1 software can be used to input the digital surface model data to perform geometric correction on the remote sensing image. The specific operation process can be: opening the remote sensing image in the ENVI5.1 software, enabling the rectification module in the ENVI5.1 software, importing the remote sensing image into the rectification module, and then importing the digital surface model data into the rectification module. The rectification module determines or finds the objects corresponding to the ground control points in the remote sensing image through the point cloud data corresponding to the ground control points in the digital surface model data, to perform geometric accurate matching on the remote sensing image, and correct the remote sensing image, until the matching accuracy of the point cloud data corresponding to the ground control points in the digital surface model data in the remote sensing image to find the objects corresponding to the ground control points is ensured to be a preset threshold. Generally, the preset threshold can be set to 99%, so that the geometric error of the remote sensing image can be controlled within one pixel.

[0105] It is worth noting that due to the imaging rules of remote sensing technology, the data collected by satellites, remote sensors or other possible sensors or the remote sensing image relative to the actual data in the target area will be geometrically deformed due to the curvature of the earth, changes in the terrain in the target area, and air refraction and a series of interference. Therefore, it is necessary to perform geometric correction on the remote sensing image to eliminate the errors caused by the imaging rules of remote sensing technology.

[0106] In this way, the errors caused by the imaging rules of remote sensing technology to the remote sensing image can be eliminated, and the accuracy and reliability of the remote sensing image segmentation method can be improved.

[0107] Step 1006: Fuse the digital surface model data and the corrected image to obtain the fused image.

[0108] It is worth noting that since only the remote sensing image is geometrically corrected, the corrected image and the remote sensing image include the same information, that is, the corrected image also includes the color information of each object in the target area, and therefore, the fusion image is obtained by fusing the corrected image and the point cloud data, so that the fusion image recording the color information and the height information of each object in the target area is obtained, which facilitates the segmentation processing of the fusion image.

[0109] In a possible implementation manner, referring to Figure 3 The point cloud data is converted into digital surface model data, including:

[0110] Step 1007: performing coordinate conversion on the point cloud data to obtain converted point cloud data.

[0111] Optionally, the coordinate conversion on the point cloud data can mean that the three-dimensional coordinate system of the point cloud data is unified with the three-dimensional coordinate system of the remote sensing image, and each three-dimensional coordinate in the point cloud data is converted into the three-dimensional coordinate system of the remote sensing image.

[0112] Optionally, the converted point cloud data can be digital elevation model (DEM) data.

[0113] It is worth noting that since different coordinate systems can be used when collecting the point cloud data and the remote sensing image, a point in the target area can correspond to two completely different coordinates in the point cloud data and the remote sensing image, which can increase the difficulty of subsequent geometric correction of the remote sensing image. The point cloud data is converted to obtain converted point cloud data, which can unify the coordinates corresponding to each point in the target area in the point cloud data and the remote sensing image, can improve the reliability and accuracy of the geometric correction of the remote sensing image, and can improve the segmentation effect of the remote sensing image.

[0114] Step 1008: classifying the converted point cloud data to obtain a plurality of classified point cloud sets.

[0115] Optionally, each point cloud set can include point cloud data of a height range, and each point cloud set can also include point cloud data of a type.

[0116] Optionally, the classification of the converted point cloud data can mean classifying each object in the target area indicated in the converted point cloud data.

[0117] Specifically, each feature in the converted point cloud data can be classified according to a preset type, which can include a building, a vegetation, a moving object (a car, a ship, an airplane, etc.), a ground point, a road point, a pipeline, a low point, a flying point, etc., and the embodiments of the present application do not limit this.

[0118] By way of example, each feature can be classified by the height of each point in the converted point cloud data. Specifically, at least one ground point can be extracted in the converted point cloud data, i.e., the generation of DEM data, and the extraction of each ground point can use the "below surface" algorithm, which can classify points below adjacent points in the source class. For example, a ground elevation can be selected as an elevation ground point to locate points slightly below the ground or above the ground. For each ground point, the algorithm can find 25 nearest neighbors of each ground point in the source class, and the algorithm can also fit each adjacent point of a ground point to a plane or a surface, and calculate the average elevation difference between each adjacent point and a plane.

[0119] If the elevation difference of any adjacent point is greater than the product of the average elevation difference and the limit difference multiple, the any adjacent point is classified. If any adjacent point is above the plane or less than the given tolerance value below the plane, the any adjacent point is not classified. In this way, each ground point can be classified according to the algorithm.

[0120] And the points above the ground can be classified after the ground points have been classified in advance. When classifying the points of the building, the points forming a plane on the roof of the building can be classified. Specifically, the points above the ground can be classified into a separate class so that the class contains points in the range of the elevation above the ground of the roof of the building. The algorithm can start classification from the blank area in the ground class, and can find points on the plane above the blank area.

[0121] The classification of vegetation points can use the classification logic of irregular, messy, cluster and higher than the ground, and the classification of pipeline points can use the classification logic of continuous and obviously higher than the ground. When classifying the points of the moving object, manual intervention classification can be used, i.e., manual classification. In addition, in the embodiments of the present application, each feature in the converted point cloud data can be classified by any possible algorithm, and the embodiments of the present application do not limit this.

[0122] The classification of the converted point cloud data is to classify all the points returned by the laser measurement scanning according to the real-world features, i.e., the points belonging to buildings are classified as building point clouds, the points belonging to vegetation are classified as vegetation point clouds, the points belonging to pipelines are classified as pipeline point clouds, and the like. After the classification, a DSM is generated. The DSM based on irregularly distributed sampling points is essentially a processing process of regular gridding of discrete data. The point-by-point interpolation method is usually used to directly establish the DSM.

[0123] In addition, the point-by-point interpolation method can refer to a method of determining a neighborhood range with the interpolation point as the center and using the sampling points falling within the neighborhood range to calculate the elevation value of the interpolation point. In the embodiment of the present application, the interpolation mathematical model used by the point-by-point interpolation method is a curved surface fitting model, and the basic steps are as follows:

[0124] (1) defining the neighborhood range of the interpolation point, (2) determining the sampling points falling within the neighborhood, (3) selecting an interpolation mathematical model, and (4) calculating the elevation of the interpolation point through the sampling points within the neighborhood and the interpolation mathematical model.

[0125] In this way, the converted point cloud data can be divided into point cloud sets of different types and / or different heights according to the actual positions and actual types of the features in the target area, which facilitates subsequent identification of the types of features indicated by the points in the converted point cloud data and can improve the accuracy and efficiency of fusion processing and obtaining of fused images.

[0126] Step 1009: converting the plurality of point cloud sets into the digital surface model data.

[0127] In this way, subsequent geometric correction of the remote sensing image is facilitated.

[0128] In one possible implementation manner, referring to Figure 4 The digital surface model data and the corrected image are fused to obtain the fused image, including:

[0129] Step 1010: image registration of the digital surface model data and the corrected image.

[0130] Optionally, the image registration of the digital surface model data and the corrected image can specifically refer to adjusting the range of the target area corresponding to the digital surface model data and the range of the target area indicated by the corrected image to the same range.

[0131] In this way, the range of the target area indicated by the digital surface model data can be ensured to correspond to the range of the target area indicated by the corrected image, which facilitates subsequent operations.

[0132] Step 1011: performing data correlation processing on the registered digital surface model data and the corrected image, to obtain a corresponding relationship between the pixel points of the registered digital surface model data and the corrected image.

[0133] Optionally, if the digital surface model data is in the form of a three-dimensional image, the digital surface model data can be a DSM data image, and each point cloud in the digital surface model data can be represented by a gray value of each pixel point in the DSM data image. The embodiments of the present application do not make any limitation in this regard.

[0134] Optionally, the data correlation processing on the registered digital surface model data and the corrected image can be to calculate the gray value relationship between the high-resolution image and the DSM data image, to match the gray values of the fused bands, and to adjust the gray distribution of each band, so as to reduce the color deviation of the fused image.

[0135] The gray value indicated by the digital surface model data can be used to indicate the height information of each object in the target region, and the gray value change indicated by the digital surface model data can be used to indicate the ups and downs of the terrain in the target region. The gray value indicated by the corrected image can be used to indicate the color information of each object in the target region.

[0136] For example, the gray value relationship between the high-resolution image and the DSM data image can be calculated based on the least square method, and the gray values of the fused bands can be matched by using the minimum variance. The embodiments of the present application do not make any limitation in this regard.

[0137] In this way, the corresponding relationship between the digital surface model data and the pixel points of the corrected image can be accurately determined, and subsequent operations can be facilitated.

[0138] Step 1012: according to the corresponding relationship between the pixel points of the registered digital surface model data and the corrected image, assigning the height information in the registered digital surface model data to each pixel point in the corrected image, and taking the corrected image with the assigned height information as the fused image.

[0139] It is worth noting that, since the corresponding relationship between the digital surface model data and the pixel points of the corrected image has been determined through image registration, the corresponding relationship between each point cloud in the digital surface model data and each pixel point of the corrected image can be determined. Then, by assigning the height information indicated by each point cloud in the registered digital surface model data to each pixel point corresponding to each point cloud in the corrected image, the purpose of accurately fusing the height information in the digital surface model data into the corrected image can be achieved, i.e., the purpose of fusing the remote sensing image and the point cloud data can be achieved.

[0140] In this way, the complementary information between the remote sensing image and the point cloud data can be extracted through fusion processing, thereby eliminating the low data quality defects of the remote sensing image or the point cloud data and improving the data accuracy of the fused image.

[0141] In one possible implementation, see [link to relevant documentation]. Figure 5 Based on the pixel information of each pixel in the fused image, the fused image is segmented to obtain the segmentation result, including:

[0142] Step 1013: Based on the pixel information of each pixel in the fused image, use the superpixel segmentation algorithm to segment the fused image and obtain the segmentation result.

[0143] It is worth noting that this superpixel segmentation algorithm refers to an algorithm that can group pixels based on the similarity of features among pixels in the fused image. Specifically, this superpixel segmentation algorithm can segment multiple adjacent pixels with similar texture, color, brightness, elevation information, and other features into an irregular superpixel block.

[0144] Furthermore, when segmenting the fused image for the first time using the superpixel segmentation algorithm, it can be first divided into k superpixel blocks, where k is a positive integer greater than 1. The larger the value of k, the more refined the segmentation of the fused image. The value of k can be adjusted according to actual needs, specifically determined based on the pixel values ​​of the fused image or the number of pixels in the fused image.

[0145] For example, assuming the pixel value of the fused image is 782×428, then the number of pixels in the fused image is 334696. The value of k can be set to 100, 8000, or any other positive integer greater than 1. This application embodiment does not limit this.

[0146] In this way, a small number of superpixel blocks can be used to replace a large number of pixels to express the features of the fused image, thereby reducing the complexity of remote sensing image segmentation processing.

[0147] In one possible implementation, see [link to relevant documentation]. Figure 6 Based on the pixel information of each pixel in the fused image, a superpixel segmentation algorithm is used to segment the fused image, obtaining the segmentation results, including:

[0148] Step 1014: Determine multiple initial seed centers in the fused image.

[0149] Optionally, each initial seed center is a pixel point in the fusion image. Each initial seed center can be a pixel point selected randomly in the fusion image, a pixel point selected in each superpixel block according to the shape and size of each superpixel block, or a pixel point in the fusion image designated by a related technical person, and the embodiments of the present application do not limit this.

[0150] The number of each initial seed center can be adjusted according to actual needs. Specifically, the number of each initial seed center can be determined according to the number of superpixel blocks into which the fusion image is divided by a superpixel segmentation algorithm, that is, the value of k. Generally, the number of each initial seed center can be k.

[0151] Step 1015: According to the pixel information of each pixel point in the fusion image, the pixel distance of each associated pixel point in the neighborhood corresponding to each initial seed center is determined, and each initial seed center is adjusted according to the pixel distance of each associated pixel point in the corresponding neighborhood to obtain at least one new seed center. The step is repeatedly executed until the pixel distance between each associated pixel point in the neighborhood corresponding to the current adjusted new seed center and the initial seed center meets a preset condition, and the neighborhood corresponding to the current adjusted new seed center is taken as a segmentation region.

[0152] Optionally, the neighborhood corresponding to each initial seed center can be a region in the fusion image in the superpixel block in which each initial seed center is located. Generally, each initial seed center corresponds to one neighborhood, and only one initial seed center is in one neighborhood.

[0153] Illustratively, the neighborhood corresponding to each initial seed center can be a nearby region with each initial seed center as the center. The size of the nearby region can be set according to actual needs. Generally, the size of the nearby region can be set to 10x10, where 10 can mean ten pixel points.

[0154] Each initial seed center and each associated pixel point of each initial seed center can be included in the neighborhood corresponding to each initial seed center. That is, each initial seed center can be a clustering center of each associated pixel point in the corresponding neighborhood.

[0155] Each associated pixel point can be another pixel point located in the same nearby region as each initial seed center.

[0156] Optionally, each new seed center can be one pixel point in each associated pixel point. Specifically, each new seed center can be one pixel point in a neighborhood corresponding to each initial seed center and having a minimum distance value to all pixel points. In this way, each new seed center can be prevented from being located at an edge position of each neighborhood, and the fused image can be more accurately segmented into fine segmentation regions, thereby improving the segmentation effect of the remote sensing image.

[0157] In addition, the associated pixel points corresponding to each new seed center can be re-divided according to the distance of each pixel point in the fused image to each new seed center, that is, the neighborhood corresponding to each new seed center can be re-divided.

[0158] The neighborhood corresponding to each new seed center can be a region in the superpixel block where each new seed center is located in the fused image. Specifically, the neighborhood corresponding to each new seed center can be a nearby region centered on each new seed center.

[0159] Generally, the distance between each new seed center and other pixel points in the neighborhood corresponding to the new seed center can be the same.

[0160] The size of the neighborhood corresponding to each new seed center can be the same as or different from the size of the neighborhood corresponding to each initial seed center, which is not limited in the embodiments of the present application.

[0161] Optionally, the preset condition can be any possible condition set by a relevant technical person according to actual needs. Generally, the preset condition can be that the pixel distance between each associated pixel point in the neighborhood corresponding to the new seed center and the initial seed center is the same as the pixel distance between each associated pixel point in the neighborhood corresponding to the new seed center and the new seed center after the last adjustment, or the pixel distance between each associated pixel point in the neighborhood corresponding to the new seed center and the new seed center is less than or equal to a preset pixel distance. The embodiments of the present application are not limited in this regard.

[0162] Optionally, each segmentation region can be a small region composed of pixel points that are adjacent in position and similar or identical in color, brightness, texture, elevation information, and the like.

[0163] It is worth noting that after adjusting each initial seed center to obtain at least one new seed center, when step 1015 is repeatedly executed subsequently, each new seed center can be determined according to the pixel information of each pixel point in the fused image, the pixel distance of each associated pixel point in the neighborhood corresponding to each new seed center can be determined, and each new seed center can be adjusted according to the pixel distance of each associated pixel point in the corresponding neighborhood to obtain at least one new seed center again. In this way, the fused image can be gradually segmented into more accurate segmentation regions.

[0164] It should be noted that after adjusting each initial seed center to obtain at least one new seed center, if a pixel point in the fused image is relatively close to multiple new seed centers, then the pixel point is taken as an associated pixel point of the new seed center closest to the pixel point, that is, the pixel point is divided into a neighborhood of the new seed center closest to the pixel point.

[0165] It should be noted that after each new seed center is determined, the superpixel blocks are re-divided according to each new seed center, so that more accurate new seed centers can be obtained on the basis of dividing the fused image according to each initial seed center, and the fused image is gradually divided into more accurate pixel blocks according to each new seed center, so as to achieve the purpose of dividing the fused image, thereby improving the remote sensing image segmentation effect and reducing the complexity of remote sensing image segmentation processing.

[0166] In one possible manner, when step 1015 is repeatedly performed to obtain each division region, the neighborhood corresponding to each new seed center after the current adjustment can be taken as a division region respectively, and the repeated execution of step 1015 can be stopped, when the repeated execution of step 1015 meets a preset number of times.

[0167] In one possible manner, when step 1015 is repeatedly performed to obtain each division region, the neighborhood corresponding to each new seed center after the current adjustment can be taken as a division region respectively, and the repeated execution of step 1015 can be stopped, when the new seed center after the current adjustment is completely the same as the new seed center after the last adjustment.

[0168] In this way, the flexibility of obtaining each division region and remote sensing image segmentation processing can be improved.

[0169] In one possible implementation, the pixel distance of each associated pixel point in the neighborhood corresponding to each initial seed center is determined according to the pixel information of each pixel point in the fused image, including:

[0170] The spatial distance between the initial seed center and each associated pixel point is determined according to the elevation information of the initial seed center and the elevation information of each associated pixel point associated with the position of the initial seed center.

[0171] For example, the spatial distance between the initial seed center and each associated pixel point can be determined according to the following formula (1).

[0172]

[0173] wherein d hrepresents the spatial distance between the initial seed center and each associated pixel point, i represents the initial seed center, j represents any associated pixel point, x represents the horizontal coordinate of the pixel point, y represents the vertical coordinate of the pixel point, and G represents the elevation information of the pixel point.

[0174] In this way, the spatial distance between the initial seed center and each associated pixel point can be accurately calculated, facilitating the subsequent steps.

[0175] According to the color information of the initial seed center and the color information of each associated pixel point associated with the position of the initial seed center, the color distance between the initial seed center and each associated pixel point is determined.

[0176] For example, the color distance between the initial seed center and each associated pixel point can be determined in the following formula (2).

[0177]

[0178] wherein d k represents the color distance between the initial seed center and each associated pixel point, L, a, and b represent three channels of the Lab mode respectively. Specifically, L specifically represents the brightness of the pixel point, the value range of L can be greater than or equal to 0 and less than or equal to 100, L = 0 represents the lowest brightness of the pixel point, which is black, and L = 100 represents the highest brightness of the pixel point, which is white. a represents the color of the pixel point in the range between red and green, a negative value represents that the color of the pixel point is biased towards green, and a positive value represents that the color of the pixel point is biased towards red. b represents the color of the pixel point in the range between yellow and blue, a negative value represents that the color of the pixel point is biased towards blue, and a positive value represents that the color of the pixel point is biased towards yellow.

[0179] In this way, the color distance between the initial seed center and each associated pixel point can be accurately calculated, facilitating the subsequent steps.

[0180] According to the spatial distance and the color distance, the pixel distance between the initial seed center and each associated pixel point is determined.

[0181] For example, the pixel distance between the initial seed center and each associated pixel point can be determined in the following formula (3).

[0182]

[0183] Wherein, D represents the pixel distance between the initial seed center and each associated pixel point, and β and γ respectively represent the weight of color distance and spatial distance when calculating the pixel distance. Generally, for remote sensing image data, the weight of spatial distance can be set to be greater than the weight of color distance, so as to improve the accuracy of determining the segmentation region according to the pixel distance, and further improve the remote sensing image segmentation effect.

[0184] For example, β = 0.64 and γ = 0.36 can be set, and β and γ are respectively brought into the above formula (3) to obtain the following formula (4).

[0185]

[0186] In this way, the pixel distance between the initial seed center and each associated pixel point can be accurately calculated, so as to subsequently determine the segmentation region according to the pixel distance, improve the accuracy of determining the segmentation region, and further improve the remote sensing image segmentation effect.

[0187] In a possible manner, when step 1015 is executed in a loop, the spatial distance between each new seed center and each associated pixel point in the corresponding neighborhood, the color distance between each new seed center and each associated pixel point in the corresponding neighborhood, and the pixel distance between each new seed center and each associated pixel point in the corresponding neighborhood can also be respectively calculated in the manner of the above formulas (1)-(4). For details, refer to the description of the above embodiments, and the embodiments of the present application will not be described again.

[0188] The device, equipment and computer readable storage medium provided by the present application for performing the remote sensing image segmentation method are described below. For details of the implementation process and technical effects, refer to the above description, and the following description will not be repeated.

[0189] Figure 7 is a structural schematic diagram of a remote sensing image segmentation device provided by an embodiment of the present application. Referring to Figure 7 The device comprises:

[0190] The acquisition module 201 is configured to acquire remote sensing image and point cloud data of a target region.

[0191] The fusion module 202 is configured to perform fusion processing on the remote sensing image and the point cloud data to obtain a fused image.

[0192] Optionally, the fused image comprises the elevation information of each pixel point.

[0193] The segmentation module 203 is configured to segment the fused image based on the pixel information of each pixel point in the fused image to obtain a segmentation result.

[0194] Optionally, the pixel information comprises the elevation information and color information of the pixel point.

[0195] The device is configured to execute the method provided by the foregoing embodiments, and has similar implementation principles and technical effects, which will not be described here again.

[0196] The above modules can be one or more integrated circuits configured to implement the above method, such as one or more application specific integrated circuits (ASICs), or one or more microprocessors, or one or more field programmable gate arrays (FPGAs), and the like. For example, when a certain module above is implemented in the form of a processing element scheduling code, the processing element can be a general-purpose processor, such as a central processing unit (CPU) or other processor that can invoke program code. For another example, the modules can be integrated together to implement in the form of a system on a chip (SOC).

[0197] Figure 8 is a structural schematic diagram of an electronic device provided by an embodiment of the present application. Referring to Figure 8 The electronic device includes a memory 301 and a processor 302. The memory 301 stores a computer program that can run on the processor 302. When the processor 302 executes the computer program, the steps in any of the above method embodiments are implemented.

[0198] The embodiment of the present application further provides a computer readable storage medium, which stores a computer program. When the computer program is executed by a processor, the steps in any of the above method embodiments can be implemented.

[0199] Optionally, the present application also provides a program product, for example, a computer readable storage medium, which includes a program. When the program is executed by a processor, the program is used to execute any of the above remote sensing image segmentation method embodiments.

[0200] In several embodiments of the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of units is only a logical function division. Actual implementation can have another division manner. For example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed elements can be indirect coupling or communication connection through some interface, device or unit, which can be electrical, mechanical or other forms.

[0201] The units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, i.e., may be located in one place, or may be distributed to multiple network units. Part or all of the units may be selected according to actual needs to achieve the purpose of the embodiment.

[0202] In addition, each functional unit in each embodiment of the application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of hardware plus software functional unit.

[0203] The integrated unit realized in the form of software functional unit can be stored in a computer readable storage medium. The software functional unit stored in a storage medium includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor (English: processor) to execute part of the steps of the method of each embodiment of the application. The foregoing storage medium includes a U disk, a mobile hard disk, a read-only memory (English: Read-Only Memory, abbreviated as: ROM), a random access memory (English: Random Access Memory, abbreviated as: RAM), a magnetic disk or an optical disk, and various program code storage media.

[0204] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be limited by the protection scope of the claims.

[0205] The above is only a preferred embodiment of the present application, and is not used to limit the present application. For those skilled in the art, the present application can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A method for segmenting remote sensing images, characterized in that, The method comprises: acquiring remote sensing images and point cloud data of a target area; performing fusion processing on the remote sensing images and the point cloud data to obtain fused images, the fused images including elevation information of each pixel point; segmenting the fused images based on pixel information of each pixel point in the fused images to obtain a segmentation result, the pixel information including elevation information and color information of the pixel point; the segmentation of the fused images comprises: determining a plurality of initial seed centers in the fused images, each initial seed center being a pixel point in the fused images; determining spatial distances between the initial seed center and each associated pixel point associated with the position of the initial seed center according to the elevation information of the initial seed center and the elevation information of each associated pixel point; determining color distances between the initial seed center and each associated pixel point associated with the position of the initial seed center according to the color information of the initial seed center and the color information of each associated pixel point; determining pixel distances between the initial seed center and each associated pixel point according to the spatial distances and the color distances; adjusting each initial seed center according to the pixel distances of each associated pixel point in the corresponding neighborhood to obtain at least one new seed center, and repeating the step of determining the new seed center until the pixel distances between each associated pixel point in the neighborhood corresponding to the current adjusted new seed center and the initial seed center meet a preset condition, and taking the neighborhood corresponding to each current adjusted new seed center as a segmentation region respectively; wherein the spatial distances between the initial seed center and each associated pixel point are determined by the following formula: wherein, wherein, d represents the spatial distance between the initial seed center and each of the associated pixel points, i represents the initial seed center, j represents any one of the associated pixel points, x represents the horizontal coordinate of the associated pixel point, y represents the vertical coordinate of the associated pixel point, and G represents the elevation information of the associated pixel point.

2. The method of claim 1, wherein, the fusion processing of the remote sensing images and the point cloud data to obtain the fused images comprises: converting the point cloud data into digital surface model data; performing geometric correction on the remote sensing images based on the digital surface model data to obtain corrected images; performing fusion processing on the digital surface model data and the corrected images to obtain the fused images.

3. The method of claim 2, wherein, the conversion of the point cloud data into digital surface model data comprises: performing coordinate conversion on the point cloud data to obtain converted point cloud data; classifying the converted point cloud data to obtain a plurality of classified point cloud sets, each point cloud set including point cloud data of one type; converting the plurality of point cloud sets into the digital surface model data.

4. The method of claim 2, wherein, the fusion processing of the digital surface model data and the corrected images to obtain the fused images comprises: performing image registration on the digital surface model data and the corrected images; performing data association processing on the registered digital surface model data and corrected images to obtain a corresponding relationship between pixel points of the registered digital surface model data and corrected images; assigning elevation information in the registered digital surface model data to each pixel point in the corrected images according to the corresponding relationship between pixel points of the registered digital surface model data and corrected images, and taking the corrected images with assigned elevation information as the fused images.

5. A device for segmenting remote sensing images, characterized in that it comprises: The device comprises: An acquisition module for acquiring remote sensing images and point cloud data of a target area; A fusion module for performing fusion processing on the remote sensing images and the point cloud data to obtain fused images, the fused images including elevation information of each pixel point; A segmentation module for segmenting the fused images based on pixel information of each pixel point in the fused images to obtain segmentation results, the pixel information including elevation information and color information of the pixel point; The segmentation module is specifically configured to determine a plurality of initial seed centers in the fused images, each initial seed center being a pixel point in the fused images; determine spatial distances between the initial seed center and each associated pixel point associated with the position of the initial seed center according to the elevation information of the initial seed center and the elevation information of each associated pixel point; determine color distances between the initial seed center and each associated pixel point associated with the position of the initial seed center according to the color information of the initial seed center and the color information of each associated pixel point; determine pixel distances between the initial seed center and each associated pixel point according to the spatial distances and the color distances; adjust each initial seed center according to the pixel distances of each associated pixel point in the corresponding neighborhood to obtain at least one new seed center, and repeat the step of determining the new seed center until the pixel distances between each associated pixel point in the neighborhood corresponding to the current adjusted new seed center and the initial seed center satisfy a preset condition, and take the neighborhood corresponding to each current adjusted new seed center as a segmentation region respectively; wherein the spatial distances between the initial seed center and each associated pixel point are determined by the following formula: wherein, wherein d represents the spatial distance between the initial seed center and each of the associated pixel points, i represents the initial seed center, j represents any one of the associated pixel points, x represents the horizontal coordinate of the associated pixel point, y represents the vertical coordinate of the associated pixel point, and G represents the elevation information of the associated pixel point.

6. An electronic device, comprising: comprises: A processor and a memory, the memory storing machine readable instructions executable by the processor, when the electronic device is running, the processor executes the machine readable instructions to perform the steps of the remote sensing image segmentation method as claimed in any one of claims 1 to 4.

7. A computer readable storage medium characterized in that, The computer readable storage medium stores a computer program, which is run by the processor to perform the steps of the remote sensing image segmentation method as claimed in any one of claims 1 to 4.

Images