Data compression method and system for digital elevation model

Abstract

The application provides a data compression method and system of a digital elevation model, and relates to the technical field of data compression, and comprises the following steps: constructing an original matrix according to a digital elevation model to be compressed; inputting the original matrix into a high turning point identification model respectively to obtain a plurality of labeling matrices associated with different receptive fields, and obtaining a fixed point set by taking the union of a plurality of high turning points labeled in each labeling matrix; performing weight difference processing on other elements in each labeling matrix except the fixed point set according to the associated receptive field to obtain a weight difference matrix; screening other elements in each weight difference matrix according to a preset screening threshold to obtain a screening matrix containing each screened element and the fixed point set; and replacing the element values in the screening matrix with corresponding heights in the digital elevation model to obtain a data compression result of the digital elevation model. The beneficial effect is that the data amount of the digital elevation model can be reduced as much as possible on the premise of retaining the basic form and large details of the terrain.

Description

Technical Field

[0001] This invention relates to the field of data compression technology, and in particular to a data compression method and system for digital elevation models. Background Technology

[0002] A Digital Elevation Model (DEM) is a digital simulation of ground topography (i.e., a digital representation of the surface morphology of the terrain) achieved through limited terrain elevation data. It is a physical ground model that represents ground elevation using an ordered array of numerical values.

[0003] The data format of Digital Elevation Models (DEM) has the suffix .Geotiff. Essentially, it is a raster image. Each element defines the size of a raster, usually a floating-point number denoted as A. It is also a geographic information encoding matrix that defines the three-dimensional data of the raster center point, denoted as (longitude lng, latitude lng, height h). However, the height h data is usually not expressed as the actual height data, but is summarized by H-domain mapping (0,1) floating-point numbers across the entire height range. This allows the data to be encoded in a uniform way, but it results in an excessively large amount of data, limiting the application scenarios of DEM. Summary of the Invention

[0004] To address the problems existing in the prior art, this invention provides a data compression method for digital elevation models, comprising:

[0005] Step S1: Construct an original matrix based on a digital elevation model to be compressed. The elements in the original matrix are the heights of the corresponding grid points in the digital elevation model, and the coordinates of the elements are the longitude and latitude of the grid points.

[0006] Step S2: Input the original matrix into the corresponding high inflection recognition model obtained in the pre-training process to obtain multiple labeling matrices associated with different receptive fields, and take the union of the multiple high inflection points labeled in each of the labeling matrices to obtain a set of fixed points.

[0007] Step S3: Based on the associated receptive field, perform weight difference processing on the other elements in the corresponding annotation matrix, excluding each of the high inflection points in the set of fixed points, to obtain the corresponding weight difference matrix.

[0008] Step S4: Based on a preset screening threshold, filter the other elements in each weight difference matrix except for each high inflection point in the set of fixed points to obtain a screening matrix containing each screened element and the set of fixed points.

[0009] Step S5: Replace the values ​​of each filtered element in the filtering matrix and each high turning point in the fixed set with the corresponding height in the digital elevation model to obtain the data compression result of the digital elevation model.

[0010] Preferably, step S1 includes:

[0011] Step S11: Add the longitude of each grid point in the digital elevation model to a list and count the length of the list;

[0012] Step S12: Based on the length, the heights corresponding to all the grid points are split into a dictionary, and the dictionary is traversed according to the rule of traversing the longitude first and then the dimension to form the original matrix.

[0013] Preferably, in step S3, the weight difference matrix is ​​obtained by performing weight difference processing using the following formula:

[0014]

[0015] Wherein, Q (i,j) The element values ​​corresponding to longitude i and latitude j in the weight difference matrix are used to represent the element values; k is used to represent the number of rows and columns of the submatrix covered by the receptive field; A (i,j) An is used to represent the element values ​​corresponding to longitude i and latitude j in the corresponding annotation matrix; An is used to represent A in the receptive field. (i,j) The element values ​​of the surrounding elements, n is used to represent the receptive field excluding A. (i,j) The number of other elements besides the surrounding elements.

[0016] Preferably, step S4 includes:

[0017] Step S41: For each element in each weight difference matrix other than each of the high inflection points in the set of fixed points, extract the minimum element value corresponding to the coordinate position in each weight difference matrix.

[0018] Step S42: Determine whether the minimum element value at each corresponding coordinate position is greater than the filtering threshold.

[0019] If so, the smallest element value is taken as the element value of the corresponding coordinate position of the filtering matrix, and a first mark is configured accordingly, and then proceed to step S43;

[0020] If not, then configure a second marker at the corresponding coordinate position;

[0021] Step S43: Fill each coordinate position with triangular faces, and determine whether each coordinate position forming each triangular face is configured with the first marker.

[0022] If not, delete the triangular face and then proceed to step S44;

[0023] If so, retain the triangular face and then proceed to step S44;

[0024] Step S44: After deleting each of the triangular faces, the resulting hole areas are filled and then welded to obtain the screening matrix.

[0025] Preferably, the receptive field includes a first receptive field, a second receptive field, and a third receptive field, wherein the first receptive field is larger than the second receptive field, and the second receptive field is larger than the third receptive field.

[0026] The present invention also provides a data compression system for digital elevation models, which applies the above-described data compression method. The data compression system includes:

[0027] A matrix construction module is used to construct an original matrix based on a digital elevation model to be compressed. The elements in the original matrix are the heights of the corresponding grid points in the digital elevation model, and the coordinate positions of the elements are the longitude and latitude of the grid points.

[0028] The fixed point identification module, connected to the matrix construction module, is used to input the original matrix into the corresponding pre-trained high inflection recognition model to obtain multiple labeled matrices associated with different receptive domains, and to obtain a fixed point set by taking the union of the multiple high inflection points labeled in each labeled matrix.

[0029] The weight difference processing module, connected to the fixed point identification module, is used to perform weight difference processing on the other elements in the corresponding annotation matrix, excluding each of the high inflection points in the fixed point set, according to the associated receptive field to obtain the corresponding weight difference matrix.

[0030] The element filtering module, connected to the weight difference processing module, is used to filter other elements in each weight difference matrix except for each high inflection point in the set of fixed points according to a preset filtering threshold, so as to obtain a filtering matrix containing each filtered element and the set of fixed points.

[0031] The data compression module is connected to the matrix construction module and the element filtering module respectively. It is used to replace the element values ​​of each of the filtered elements in the filtering matrix and each of the high turning points in the fixed set with the corresponding heights in the digital elevation model to obtain the data compression result of the digital elevation model.

[0032] Preferably, the matrix construction module includes:

[0033] A statistical unit is used to add the longitude of each grid point in the digital elevation model to a list and to count the length of the list.

[0034] The construction unit, connected to the statistics unit, is used to split the height corresponding to all the grid points according to the length to form a dictionary, and to traverse the dictionary according to the rule of traversing the longitude first and then the dimension to form the original matrix.

[0035] Preferably, in the weight difference processing module, the weight difference matrix is ​​obtained by performing weight difference processing using the following formula:

[0036]

[0037] Wherein, Q (i,j) The values ​​of the elements corresponding to longitude i and latitude j in the weight difference matrix are used to represent the element values; n represents the size of the receptive field; k represents the number of rows and columns of the submatrix covered by the receptive field; A (i,j) An is used to represent the element values ​​corresponding to longitude i and latitude j in the corresponding annotation matrix; An is used to represent A in the receptive field. (i,j) The element values ​​of the surrounding elements.

[0038] Preferably, the element filtering module includes:

[0039] The extraction unit is used to extract the minimum element value at the corresponding coordinate position in each of the weight difference matrices, except for each of the high inflection points in the set of fixed points;

[0040] The first judgment unit, connected to the extraction unit, is used to, when the smallest element value at the corresponding coordinate position is greater than the filtering threshold, take the smallest element value as the element value at the corresponding coordinate position of the filtering matrix and configure a first mark accordingly, and when the smallest element value at the corresponding coordinate position is not greater than the filtering threshold, configure a second mark for the element value at the corresponding coordinate position.

[0041] The second judgment unit, connected to the first judgment unit, is used to fill each of the coordinate positions with triangular faces, retain the corresponding triangular face when it is determined that each coordinate position forming each of the triangular faces is configured as the first mark, and delete the triangular face when any coordinate position of the triangular face is configured as the second mark.

[0042] The hole-filling unit, connected to the second judgment unit, is used to fill the hole areas formed after deleting each of the triangular faces and then weld them to obtain the screening matrix.

[0043] Preferably, the receptive field includes a first receptive field, a second receptive field, and a third receptive field, wherein the first receptive field is larger than the second receptive field, and the second receptive field is larger than the third receptive field.

[0044] The above technical solution has the following advantages or beneficial effects: it can minimize the amount of data in the digital elevation model while preserving the basic shape and major details of the terrain, reduce the performance requirements of the corresponding application tools when applying the digital elevation model, and effectively expand the application scenarios of the digital elevation model. Attached Figure Description

[0045] Figure 1 A flowchart illustrating a data compression method for a digital elevation model is provided in a preferred embodiment of the present invention.

[0046] Figure 2 This is a schematic diagram of a sub-process of step S1 in a preferred embodiment of the present invention.

[0047] Figure 3 This is a schematic diagram of the structure of the annotation matrix in a preferred embodiment of the present invention;

[0048] Figure 4 This is a schematic diagram of a sub-process of step S4 in a preferred embodiment of the present invention;

[0049] Figure 5 This is a schematic diagram of the result of triangular face filtering based on the first and second marks of each coordinate position in a preferred embodiment of the present invention.

[0050] Figure 6 This is a schematic diagram of the hole boundary in a preferred embodiment of the present invention;

[0051] Figure 7 In a preferred embodiment of the present invention, a comparative schematic diagram of the digital elevation model before and after data compression is provided.

[0052] Figure 8 This is a schematic diagram of the structure of a data compression system for a digital elevation model, which is a preferred embodiment of the present invention. Detailed Implementation

[0053] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. The present invention is not limited to this embodiment; other embodiments that conform to the spirit of the present invention may also fall within the scope of the present invention.

[0054] In a preferred embodiment of the present invention, based on the above-mentioned problems existing in the prior art, a data compression method for digital elevation models is provided, such as... Figure 1 As shown, it includes:

[0055] Step S1: Construct an original matrix based on a digital elevation model to be compressed. The elements in the original matrix are the heights of the corresponding grid points in the digital elevation model, and the coordinates of the elements are the longitude and latitude of the grid points.

[0056] Step S2: Input the original matrix into the corresponding high inflection recognition model obtained in the pre-training process to obtain multiple labeling matrices associated with different receptive fields, and take the union of the multiple high inflection points labeled in each labeling matrix to obtain a set of fixed points.

[0057] Step S3: Based on the associated receptive field, perform weight difference processing on the other elements in the corresponding annotation matrix except for each high inflection point in the fixed point set to obtain the corresponding weight difference matrix.

[0058] Step S4: Based on a preset screening threshold, filter the elements in each weight difference matrix except for each high inflection point in the fixed point set to obtain a screening matrix containing each screened element and the fixed point set.

[0059] Step S5: Replace the values ​​of each filtered element in the filtering matrix and each high turning point in the fixed set with the corresponding heights in the digital elevation model to obtain the data compression result of the digital elevation model.

[0060] Specifically, in this embodiment, the digital elevation model to be compressed is expanded into an original matrix using a row-determinant method, and then data compression is performed based on the original matrix. Preferably, the original matrix uses the longitude of each grid point in the digital elevation model as a column, the latitude of each grid point as a column, and the height of each grid point as the element value.

[0061] More specifically, such as Figure 2 As shown, the process of forming the original matrix by expanding the digital elevation model to be compressed using a determinant is as follows, i.e., step S1 includes:

[0062] Step S11: Add the longitude of each grid point in the digital elevation model to a list and count the length of the list;

[0063] Step S12: Based on the length, split the height corresponding to all grid points to form a dictionary, and traverse the dictionary according to the rule of traversing longitude first and then latitude to form the original matrix.

[0064] Specifically, in this embodiment, taking a digital elevation model containing 100 longitudes (i.e., the length of the aforementioned list) and 100 latitudes as an example, the total number of grid points is 10,000, meaning the number of heights is 10,000. These 10,000 heights can be arranged sequentially as a queue, which needs to be split. Preferably, this queue is split according to the number of longitudes. This allows dividing the 10,000 heights into 100 parts and storing them in a dictionary. Then, the dictionary is traversed according to the rule of first traversing longitudes and then latitudes, so that each height part is a row, arranged sequentially to form the original matrix with longitude as the row, latitude as the column, and height as the element value. It is understood that the above splitting is not limited to splitting by the number of longitudes and can be customized according to requirements.

[0065] After constructing the original matrix, it is first necessary to identify the high turning points in the original matrix that represent significant changes in terrain morphology. Each high turning point is an indispensable point for representing changes in terrain morphology and should be preserved during data compression. In this embodiment, a high-turn recognition model is used to process the original matrix. The preferred high-turn recognition model is the Unet convolutional network, which includes multiple convolutional layers and upsampling layers. Specifically, a 6*6 convolutional layer is used to convolve the input original matrix to extract features; then a 3*3 convolutional layer convolves the 6*6 feature values ​​to extract features; then a 2*2 convolutional layer convolves the 3*3 feature values ​​to extract features; then a 1*1 convolutional layer convolves the 2*2 feature values ​​to extract features; then the 2*2 feature map is upsampled to the 1*1 convolutional layer to construct a feature map; then the 3*3 feature map is upsampled to the 2*2 convolutional layer to construct a feature map; then the 6*6 feature map is upsampled to the 3*3 convolutional layer to construct a feature map; finally, the upsampled feature maps are superimposed to form a full-image feature map, which is then upsampled to the 6*6 convolutional layer to construct a label matrix.

[0066] Before proceeding, the high-turn recognition model needs to be trained. This allows for direct data compression during the subsequent data compression process, improving compression efficiency. The training process for the high-turn recognition model is as follows:

[0067] Multiple uncompressed data elevation models are obtained and expanded into corresponding uncompressed matrices. High inflection points are then marked on these uncompressed matrices to obtain uncompressed labeled matrices, which serve as the training set. This allows for the training of a data elevation model that takes the uncompressed matrices as input and outputs the uncompressed labeled matrices with high inflection points. To further improve data compression, data elevation models with different receptive domains are selected for optimal results, enabling data compression processing that considers multiple receptive domains simultaneously.

[0068] After training high-turn recognition models for different receptive domains, the original matrix can be input into the high-turn model to obtain the corresponding annotation matrix. For different receptive domains, the identified high-turn points may differ due to variations in coverage area. In this embodiment, it is preferable to add the high-turn points in each annotation matrix as fixed points to a fixed-point set, and all high-turn points in the fixed-point set are retained during subsequent compression processing.

[0069] As can be seen, the high inflection points identified above can only represent significant changes in terrain morphology. However, terrain conditions are usually quite complex. For relatively flat areas, fewer points are typically used for representation. For areas where the terrain morphology does not change significantly but is not flat, more points are needed to ensure that the compressed data elevation model still retains the basic terrain morphology and major details. Based on this, after identifying the high inflection points that must be retained, it is necessary to further extract other points that need to be retained. In this embodiment, a weighted difference processing method combined with a screening threshold is used to filter out other points that need to be retained from the annotation matrices corresponding to each receptive field.

[0070] In a preferred embodiment of the present invention, the weight difference matrix is ​​obtained by weight difference processing using the following formula:

[0071]

[0072] Among them, Q (i,j) The element values ​​corresponding to longitude i and latitude j in the weight difference matrix are used to represent the element values; k is used to represent the number of rows and columns of the submatrix covered by the receptive field; A (i,j) An is used to represent the element values ​​corresponding to longitude i and latitude j in the corresponding labeling matrix; An is used to represent A in the receptive field. (i,j) The element values ​​of the surrounding elements, n is used to represent the receptive field excluding A. (i,j) The number of other elements besides the surrounding elements.

[0073] Specifically, in this embodiment, as follows Figure 3 The labeled matrix shown in the figure (the high inflection points of the labels are not shown in the figure, but are only used as an example to illustrate how to perform weight difference processing) is used as an example. (1,1) For H1,A (20,20) For H400, the others follow the same logic. Taking a receptive field size of 3*3 as an example, for A... (2,2) ,like Figure 3 As shown, in a 3*3 receptive field, A with element value H22 (2,2) The element value Q in the weight difference matrix (2,2) The following can be calculated:

[0074]

[0075] Right now

[0076] The calculation methods for other receptive domains are based on this example and will not be elaborated here. It is worth noting that for element values ​​in the edge region, such as H1, there are missing elements in its 3*3 receptive domain. In this embodiment, it is preferable to expand outwards by one layer from the outermost ring so that the element values ​​at the edge can participate in the calculation. The element values ​​of the expanded layer are the same as the element values ​​of the outermost ring.

[0077] After processing the label matrix corresponding to each receptive domain to obtain the corresponding weight difference matrix, in a preferred embodiment of the present invention, as follows: Figure 4 As shown, step S4 includes:

[0078] Step S41: For each element in each weight difference matrix other than each high inflection point in the set of fixed points, extract the minimum element value at the corresponding coordinate position in each weight difference matrix.

[0079] Step S42: Determine whether the minimum element value at each corresponding coordinate position is greater than the filtering threshold.

[0080] If so, the smallest element value is used as the element value at the corresponding coordinate position of the filtering matrix, and a first mark is configured accordingly, and then proceed to step S43;

[0081] If not, then configure a second tag for the element value at the corresponding coordinate position;

[0082] Step S43: Fill each coordinate position with triangular faces, and determine whether each coordinate position forming each triangular face is configured as the first marker:

[0083] If not, delete the triangular face and then proceed to step S44;

[0084] If so, retain the triangular face and then proceed to step S44;

[0085] Step S44: After deleting each triangular face, fill the hole area formed by the hole and then weld it to obtain the screening matrix.

[0086] Specifically, in this embodiment, taking three receptive domains as an example, after processing the label matrix corresponding to each receptive domain to obtain the corresponding weight difference matrix, for the above A... (2,2) The coordinates of the three receptive domains are selected from Q. (2,2) The smallest element value, and then determine Q. (2,2) If the minimum element value is greater than the filtering threshold, the point is retained and can be assigned a first marker with a value of 1. If the minimum element value is not greater than the threshold, the point is assigned a second marker with a value of 0.

[0087] After marking the points to be deleted, the subsequent deletion operation is required. Specifically, firstly, triangular faces are filled based on each coordinate position, using the annotation matrix as an example. Figure 3 Taking a 20*20 area as an example, the coordinate positions for triangle filling are also 20*20. The preferred method for triangle filling is as follows: for the four coordinate positions (i, j), (i+1, j), (i, j+1), and (i+1, j+1), two triangles can be formed. One triangle is formed by (i, j), (i+1, j), and (i, j+1), and the other triangle is formed by (i+1, j), (i, j+1), and (i+1, j+1). The above triangle filling method is only one embodiment and is not intended to limit the scope of this technical solution.

[0088] After forming the aforementioned triangular faces, the triangular faces are filtered based on the first and second marks of each coordinate position. Specifically, if the coordinate position of any vertex of a triangular face matches the second mark, then that triangular face is retained only if the coordinate positions of all three vertices of that triangular face match the first mark. Figure 5 As shown, the dashed lines represent the triangular faces to be deleted, and the solid lines represent the triangular faces to be retained. It can be seen that if the triangular faces represented by the dashed lines are deleted, multiple hole areas will be formed. In order to ensure the continuity of the digital elevation model after data compression, the hole areas need to be filled after deletion.

[0089] Specifically, clustering can be performed on points at each coordinate location with the second label, such as for... Figure 6 The area of ​​holes shown ( Figure 5 The clustering result for the hole area in the upper left corner is as follows: Figure 6 The thick lines shown connect the four points forming a rectangle. Based on the coordinates of these four points, expand outwards by subtracting 1 from the smallest coordinate value and adding 1 to the largest coordinate value, resulting in the rectangle shown. Figure 6 The hexagonal boundary shown in the thick line represents the boundary of the hole in the area. Based on this boundary, a triangulation algorithm is preferably used to fill the hole. After filling the hole, to avoid breakage at the hole boundary, the Weld algorithm is preferably used for welding.

[0090] As a further preferred option, the uncompressed matrix can be weighted before labeling, so that model trainers can more intuitively see the changes in terrain morphology, which will facilitate labeling. The specific processing method is the same as above, and will not be repeated here.

[0091] The resulting filtering matrix contains all the points that need to be retained. As shown in the above processing, the element values ​​in the filtering matrix are not the original heights. Therefore, it is necessary to replace the element values ​​in the filtering matrix to obtain the compressed data result of the digital elevation model. Figure 7 As shown, the left side is the uncompressed digital elevation model, and the right side is the compressed digital elevation model data.

[0092] In a preferred embodiment of the present invention, the receptive field includes a first receptive field, a second receptive field, and a third receptive field, wherein the first receptive field is larger than the second receptive field, and the second receptive field is larger than the third receptive field.

[0093] Specifically, in this embodiment, the preferred size of the first receptive field is 3*3, the size of the second receptive field is 5*5, and the size of the third receptive field is 7*7. The number and size of each receptive field are not limited thereto and can be customized according to requirements.

[0094] This invention also provides a data compression system for digital elevation models, applying the above-described data compression method, such as... Figure 8 As shown, the data compression system includes:

[0095] Matrix construction module 1 is used to construct an original matrix based on a digital elevation model to be compressed. The elements in the original matrix are the heights of the corresponding grid points in the digital elevation model, and the coordinates of the elements are the longitude and latitude of the grid points.

[0096] The fixed point recognition module 2 and the connection matrix construction module 1 are used to input the original matrix into the corresponding high inflection recognition model obtained in the pre-training process to obtain multiple labeled matrices associated with different receptive fields, and to obtain a set of fixed points by taking the union of the multiple high inflection points labeled in each labeled matrix.

[0097] The weight difference processing module 3 is connected to the fixed point identification module 2. It is used to perform weight difference processing on the other elements in the corresponding annotation matrix, except for each high inflection point in the fixed point set, according to the associated receptive field to obtain the corresponding weight difference matrix.

[0098] The element filtering module 4 is connected to the weight difference processing module 3. It is used to filter the other elements in each weight difference matrix except for each high inflection point in the set of fixed points according to a preset filtering threshold, so as to obtain a filtering matrix containing each filtered element and the set of fixed points.

[0099] The data compression module 5 is connected to the matrix construction module 1 and the element filtering module 4 respectively. It is used to replace the element values ​​of each filtered element in the filtering matrix and each high turning point in the fixed set with the corresponding height in the digital elevation model, so as to obtain the data compression result of the digital elevation model.

[0100] In a preferred embodiment of the present invention, the matrix construction module 1 includes:

[0101] The statistics unit 11 is used to add the longitude of each grid point in the digital elevation model to a list and count the length of the list.

[0102] The construction unit 12 is connected to the statistics unit 11. It is used to split the height corresponding to all grid points according to the length to form a dictionary, and to traverse the dictionary according to the rule of traversing longitude first and then traversing dimension to form the original matrix.

[0103] In a preferred embodiment of the present invention, the weight difference processing module 3 uses the following formula to perform weight difference processing to obtain the weight difference matrix:

[0104]

[0105] Among them, Q (i,j) The element values ​​corresponding to longitude i and latitude j in the weight difference matrix are used to represent the element values; n represents the size of the receptive field; k represents the number of rows and columns of the submatrix covered by the receptive field; A (i,j) An is used to represent the element values ​​corresponding to longitude i and latitude j in the corresponding labeling matrix; An is used to represent A in the receptive field. (i,j) The element values ​​of the surrounding elements.

[0106] In a preferred embodiment of the present invention, the element filtering module 4 includes:

[0107] Extraction unit 41 is used to extract the minimum element value at the corresponding coordinate position in each weight difference matrix for each element other than each high inflection point in the set of fixed points.

[0108] The first judgment unit 42 is connected to the extraction unit 41. When the smallest element value at the corresponding coordinate position is greater than the filtering threshold, the smallest element value is used as the element value at the corresponding coordinate position of the filtering matrix and a first mark is configured accordingly. When the smallest element value at the corresponding coordinate position is not greater than the filtering threshold, the element value at the corresponding coordinate position is configured with a second mark.

[0109] The second judgment unit 43 is connected to the first judgment unit 42 and is used to fill each coordinate position with triangular faces. When each coordinate position forming each triangular face is configured as the first mark, the corresponding triangular face is retained, and when any coordinate position of the triangular face is configured as the second mark, the triangular face is deleted.

[0110] The hole-filling unit 44 is connected to the second judgment unit 43 and is used to fill the hole areas formed after deleting each triangular face and then weld them to obtain the screening matrix.

[0111] In a preferred embodiment of the present invention, the receptive field includes a first receptive field, a second receptive field, and a third receptive field, wherein the first receptive field is larger than the second receptive field, and the second receptive field is larger than the third receptive field.

[0112] The above description is merely a preferred embodiment of the present invention and does not limit the implementation and protection scope of the present invention. Those skilled in the art should realize that any equivalent substitutions and obvious changes made using the content of this specification and illustrations should be included within the protection scope of the present invention.

Claims

1. A data compression method for digital elevation models, characterized in that, include: Step S1: Construct an original matrix based on a digital elevation model to be compressed. The elements in the original matrix are the heights of the corresponding grid points in the digital elevation model, and the coordinates of the elements are the longitude and latitude of the grid points. Step S2: Input the original matrix into the corresponding high inflection recognition model obtained in the pre-training process to obtain multiple labeling matrices associated with different receptive domains, and take the union of the multiple high inflection points labeled in each labeling matrix to obtain a set of fixed points. Step S3: Based on the associated receptive field, the elements in the corresponding annotation matrix other than the high inflection points in the fixed point set are weighted using the following formula to obtain the corresponding weight difference matrix: ； Among them, the Used to represent the longitude in the weight difference matrix Dimensions The element value of the corresponding element; Used to represent the number of rows and columns of the submatrix covered by the receptive field; Used to represent the longitude in the corresponding label matrix Dimensions The element value of the corresponding element; Used to represent the receptive domain The element values ​​of the surrounding elements, Used to indicate that, except for the receptive field The number of elements other than the surrounding elements; Step S4: Based on a preset screening threshold, filter the other elements in each weight difference matrix except for each high inflection point in the set of fixed points to obtain a screening matrix containing each screened element and the set of fixed points. Step S5: Replace the element values ​​of each filtered element in the filtering matrix and each high turning point in the set of fixed points with the corresponding height in the digital elevation model to obtain the data compression result of the digital elevation model.

2. The data compression method according to claim 1, characterized in that, Step S1 includes: Step S11: Add the longitude of each grid point in the digital elevation model to a list and count the length of the list; Step S12: Based on the length, the heights corresponding to all the grid points are split into a dictionary, and the dictionary is traversed according to the rule of traversing the longitude first and then the dimension to form the original matrix.

3. The data compression method according to claim 1, characterized in that, Step S4 includes: Step S41: For each element in each weight difference matrix other than each of the high inflection points in the set of fixed points, extract the minimum element value corresponding to the coordinate position in each weight difference matrix. Step S42: Determine whether the minimum element value at each corresponding coordinate position is greater than the filtering threshold. If so, the smallest element value is taken as the element value of the corresponding coordinate position of the filtering matrix, and a first mark is configured accordingly, and then proceed to step S43; If not, then configure a second tag for the element value at the corresponding coordinate position; Step S43: Fill each coordinate position with triangular faces, and determine whether each coordinate position forming each triangular face is configured with the first marker. If not, delete the triangular face and then proceed to step S44; If so, retain the triangular face and then proceed to step S44; Step S44: After deleting each of the triangular faces, the resulting hole areas are filled and then welded to obtain the screening matrix.

4. The data compression method according to claim 1, characterized in that, The receptive field includes a first receptive field, a second receptive field, and a third receptive field, wherein the first receptive field is larger than the second receptive field, and the second receptive field is larger than the third receptive field.

5. A data compression system for a digital elevation model, characterized in that, The data compression system, employing the data compression method as described in any one of claims 1-4, comprises: A matrix construction module is used to construct an original matrix based on a digital elevation model to be compressed. The elements in the original matrix are the heights of the corresponding grid points in the digital elevation model, and the coordinate positions of the elements are the longitude and latitude of the grid points. The fixed point identification module, connected to the matrix construction module, is used to input the original matrix into the corresponding pre-trained high inflection recognition model to obtain multiple labeled matrices associated with different receptive domains, and to obtain a fixed point set by taking the union of the multiple high inflection points labeled in each labeled matrix. The weight difference processing module, connected to the fixed point identification module, is used to perform weight difference processing on the elements in the corresponding annotation matrix, excluding each high inflection point in the fixed point set, according to the associated receptive field, to obtain the corresponding weight difference matrix: ； Among them, the Used to represent the longitude in the weight difference matrix Dimensions The element value of the corresponding element; Used to represent the number of rows and columns of the submatrix covered by the receptive field; Used to represent the longitude in the corresponding label matrix Dimensions The element value of the corresponding element; Used to represent the receptive field The element values ​​of the surrounding elements, Used to indicate that, except for the receptive field The number of elements other than the surrounding elements; The element filtering module, connected to the weight difference processing module, is used to filter other elements in each weight difference matrix except for each high inflection point in the set of fixed points according to a preset filtering threshold, so as to obtain a filtering matrix containing each filtered element and the set of fixed points. The data compression module is connected to the matrix construction module and the element filtering module respectively. It is used to replace the element values ​​of each of the filtered elements in the filtering matrix and each of the high turning points in the set of fixed points with the corresponding heights in the digital elevation model, so as to obtain the data compression result of the digital elevation model.

6. The data compression system according to claim 5, characterized in that, The matrix construction module includes: A statistical unit is used to add the longitude of each grid point in the digital elevation model to a list and to count the length of the list. The construction unit, connected to the statistics unit, is used to split the height corresponding to all the grid points according to the length to form a dictionary, and to traverse the dictionary according to the rule of traversing the longitude first and then the dimension to form the original matrix.

7. The data compression system according to claim 5, characterized in that, The element filtering module includes: The extraction unit is used to extract the minimum element value at the corresponding coordinate position in each of the weight difference matrices, except for each of the high inflection points in the set of fixed points; The first judgment unit, connected to the extraction unit, is used to, when the smallest element value at the corresponding coordinate position is greater than the filtering threshold, take the smallest element value as the element value at the corresponding coordinate position of the filtering matrix and configure a first mark accordingly, and when the smallest element value at the corresponding coordinate position is not greater than the filtering threshold, configure a second mark for the element value at the corresponding coordinate position. The second judgment unit, connected to the first judgment unit, is used to fill each of the coordinate positions with triangular faces, retain the corresponding triangular face when it is determined that each coordinate position forming each of the triangular faces is configured as the first mark, and delete the triangular face when any coordinate position of the triangular face is configured as the second mark. The hole-filling unit, connected to the second judgment unit, is used to fill the hole areas formed after deleting each of the triangular faces and then weld them to obtain the screening matrix.

8. The data compression system according to claim 5, characterized in that, The receptive field includes a first receptive field, a second receptive field, and a third receptive field, wherein the first receptive field is larger than the second receptive field, and the second receptive field is larger than the third receptive field.

Images