CN110162650B - Small image spot melting method considering local optimization and overall area balance - Google Patents

Abstract

The embodiment of the present invention discloses a small patch fusion method that takes into account both local optimum and overall area balance, including: A. Obtaining full overlay vector image patch data of an image to be processed; and obtaining adjacent images in it according to the data The data information of the spot; B. According to the data information of the adjacent spots, the area of the small spots in the image is pre-allocated, and the subdivision area of each adjacent spot to the small spots is obtained; C. After the area is pre-allocated Count the areas of the first-level land types of the small map spots, and calculate the respective change rates of the area of each type before and after the pre-allocation; D. When the respective change rates are lower than the second specified threshold, determine the small map spots According to the skeleton line, each small patch is split into multiple fragments; the fragments are merged into adjacent patches, so that the small patches are melted to form the merged full-coverage patch data. From the above, the present application can effectively maintain the consistency of the land types before and after the patch merger.

Description

A small patch melting method that takes into account both local optimum and overall area balance

technical field

The invention relates to the field of map mapping and map synthesis, in particular to a small map spot melting method that takes into account both local optimum and overall area balance.

Background technique

With the increasing depth and breadth of the application of geographic and national census data, the need to comprehensively reflect the regional land use status at multiple levels from large-scale land cover data (plots) to various small-scale data is becoming more and more urgent. However, the land use thematic data is a full-coverage, non-overlapping spatial overlay. When the map changes from a large scale to a small scale, the small patches must be melted because the area is too small to continue to be expressed on the map (Ai Tinghua et al., 2010). The basic idea of the Dissolving operation is to split the small patch into several "fragments" according to certain rules according to the conditions of the adjacent polygon elements, and merge these "fragments" into the adjacent patches, so that the elements in the map frame are more compact. concise. Among them, the small patches refer to the discretely distributed fine surface cover data in the land use data, which are small in area, but large in number, complex and diverse in shape and widely distributed in the region. When synthesizing land use data, the melting results of small patches can directly affect the quality of land use data synthesis results.

The key to the fusion of patch splitting is how to delineate the splitting line in the small patch. Delaunay triangulation has the advantages of "circle rule" or "maximum and minimum angle rule", and has become a common method for patch fusion synthesis (Ware et al, 1997a). Jones et al. (1995) proposed a split line extraction method based on Delaunay triangulation, which uses triangulation to subdivide the interior of small blobs and dissolve the small blobs. Ai Tinghua et al. (2000) proposed a synthesis problem for polygonal surface objects, which can realize polygon merging and other operations by establishing a constrained Delaunay triangulation structure. Gao et al. (2004) combined with thematic knowledge, based on the Delaunay triangulation to extract small patch split lines to perform dimensionality reduction (Collapse) operation on land use data. However, the above operations are all based on the central axis division of the triangulation. It is not conducive to maintain the percentage of land area of various types before and after integration. The splitting lines of small patches should be adjusted according to the importance of adjacent patches. To this end, a weighted skeleton line segmentation strategy is proposed. Liu Yaolin et al. (2010) used this method to establish an improved algorithm for constrained Delaunay triangulation with weighted skeleton lines considering spatial proximity and semantic proximity, and the comprehensive results kept the characteristics of land use better. Similarly, Meijers et al. proposed the SPLITAREA method to achieve the weighted split line extraction of the triangulation based on local elements.

However, the existing algorithms mostly consider the local optimal constraints, which leads to the shortcomings of the fusion results of small patches: for example, the similarity between small-area patches and adjacent patches and the length of the shared boundary are regarded as the local optimum, which leads to the globalization after subdivision and merger. The area of the land type varies greatly, and the output result does not conform to the actual situation. For example, Cheng and Li (2006) used two local optimal calculation methods of adjacent patch area and shared boundary length to dissolve small patches, and found that 12.3% and 8.3% of land types changed before and after the merger, respectively. .

Therefore, there is an urgent need for a small patch fusion method that takes into account both the local optimum and the overall area balance, so as to solve or partially solve the above technical problems, so as to better maintain the consistency of the land types before and after the merger of the patches.

SUMMARY OF THE INVENTION

In view of this, the main purpose of the present invention is to propose a small patch melting method that takes into account both local optimum and overall area balance, so as to better maintain the consistency of land types before and after patch merger.

A patch fusion method that takes into account both local optimum and overall area balance provided by the present application includes:

A. Obtain full-coverage vector image spot data of an image to be processed; and obtain data information of adjacent spots in it according to the data; wherein, the data information of adjacent spots includes at least one of the following: Area, shared edge length and semantic distance;

B. According to the data information of the adjacent spots, the area of the small spots in the image is pre-allocated, and the subdivision area of each adjacent spot to the small spots is obtained; wherein, the area is smaller than the first specified threshold. spots as small pattern spots;

C. Count the area of the first-level land type of each of the small patches after the area pre-allocation, and calculate the respective change rates of the area of each type before and after the pre-allocation;

D. When the respective change rates are lower than the second specified threshold, determine the internal skeleton line of each of the small image spots, and split each small image spot into a plurality of fragments according to the skeleton lines; and The fragments are merged into adjacent patches, so that the small patches are melted to form the merged full-coverage patch data.

From the above, the patch melting method proposed in the present application can maintain the balance of the overall land type area before and after melting while taking into account the optimal local spatial pattern of the patch during the melting process. It is beneficial to better maintain the consistency of the land types before and after the merger of the map patches.

Among them, the first-level land type is the distribution standard of the first-level land type of the national standard, including: cultivated land, forest land, grassland, water area, (urban and rural, working conditions, residents) land, unused land.

Preferably, the step C also includes:

When there is a rate of change higher than the second specified threshold in each of the rates of change, the difference between all land types and the second specified threshold is recorded, and the area pre-segmentation result is iteratively processed according to the area balance iterative algorithm to Each of the rates of change is below the second specified threshold.

From the above, it is conducive to better allocation and further melting processing.

Preferably, the calculation model for obtaining the area of the adjacent patch in step A is:

Among them, i is the clockwise number of each node of the adjacent polygon, _xi is the abscissa of each node of the adjacent polygon, and y _i+1 and y _i-1 are the ordinate of each node of the adjacent polygon.

From the above, it is beneficial to obtain the area of adjacent patches more accurately and effectively.

Preferably, the calculation model for obtaining the length of the shared edge in step A is:

Wherein, x _i and y _i are respectively the abscissa and ordinate of an end point of the shared side; x _i+1 and y _i+1 respectively represent the abscissa and ordinate of the other end of the shared side.

From the above, it is beneficial to obtain the shared edge length more accurately and effectively.

Preferably, the step B includes:

B1. Obtain the subdivision capability of each adjacent dot of each small dot for obtaining the subdivision area of the small dot;

B2. According to the ratio of the segmenting capability, obtain the segmented area of each adjacent patch of the small patch.

From the above, it is beneficial to obtain the subdivision area of each adjacent patch of the small patch more accurately and effectively.

Preferably, the calculation model of the dissection capability described in step B1 is:

Among them, i=1, 2, 3 in S _i , S ₁ , S ₂ , and S ₃ represent three constraint indicators of adjacent patch area, shared edge length, and semantic distance, respectively, and _wi is the weight of each indicator; , a represents the small patch, and b _i represents the i-th adjacent patch.

From the above, it is beneficial to obtain the segmentation capability of each adjacent patch more accurately and effectively.

Preferably, in the step B2, according to the ratio of the subdivision capability, the calculation model for obtaining the subdivision area of each adjacent dot to the small dot is:

Area _i =SAF(a, b _i )/SAF(a, b)*Area

Among them, Area _i is the subdivision area of the i-th adjacent dot to the small dot; SAF(a,b) is the sum of the subdivision capabilities of all adjacent dots; Area is the area of the small dot; SAF(a,b) _i ) represents the ability of the i-th adjacent patch to obtain the segmented area of the small patch.

From the above, it is beneficial to obtain the subdivision area of each adjacent patch to the small patch more accurately and effectively.

Preferably, the step D includes:

D1, according to the Delaunay triangulation method, determine the adjacent image spots that form a subdivision for the small image spots;

D2. Perform pairwise calculations on the small image spots and the adjacent image spots, respectively, to obtain subdivision points;

D3. Generate a split line according to the split point, and split the small patch into a plurality of fragments accordingly, and merge the fragments into adjacent patches, so that the small patches are melted to The full-coverage patch data after the merger is formed.

From the above, it is conducive to more reasonable melting of small patches to form full overlay patch data after the merger.

Preferably, the calculation formula for obtaining the division points is:

Among them, a represents the small patch, b represents a neighboring patch, c represents another neighboring patch, (x _b , y _b ), (x _c , y _c ), respectively represent the two ends of the edge in the triangular network Coordinates, where x _b and x _c represent the abscissa, and y _b and y _c represent the ordinate; Area(a,b), Area(a,c) are the divisions of adjacent patches b and c to small patch a, respectively area value.

From the above, it is beneficial to obtain the subdivision points more accurately and effectively.

By means of the above scheme, the present invention has at least the following advantages:

The patch melting method proposed in the present application, while taking into account the optimal local spatial pattern of the patch during the melting process, can maintain the balance of the overall land type area before and after the melting. It is beneficial to better maintain the consistency of the land types before and after the merger of the map patches.

Description of drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the accompanying drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description These are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can also be obtained according to these drawings without creative efforts.

Figure 1 is a schematic diagram of an existing small patch fusion method taking into account the competitive segmentation ability of adjacent patch spaces: (a) the original skeleton line, (b) the adjusted skeleton line;

Figure 2 is a schematic diagram of the shortcomings of the existing small patch melting methods considering local optimization: (a) the original land use data and the small patches (deepened part) in it, (b) based on Delaunay triangulation using local optimization The split line proposed by the algorithm (bold line), (c) the melting result;

Fig. 3a, b are the flow charts of the method for dissolving small image spots that take into account local optimum and overall area balance provided by the present invention;

4 is a schematic diagram of the adjustment value statistics of the area change rates of various categories made according to an embodiment of the present invention;

FIG. 5 is a schematic diagram of iterative adjustment of the dissolving area of a small patch taking into account the global optimum made by an embodiment of the present invention;

FIG. 6 is a schematic diagram of the area statistics of the patch adjustment of grassland and artificial structures made according to an embodiment of the present invention;

FIG. 7 is a schematic diagram of a small map spot melting result of land use pavement data made in an embodiment of the present invention;

FIG. 8 is a schematic diagram showing the comparison of the melting area of the small map spot to the surrounding land type map spot according to the embodiment of the present invention.

Detailed ways

Example 1

Referring to the accompanying drawings, a method for merging patches provided by the present invention for maintaining the contour features of structured objects will be described in detail.

As shown in Figure 1, it is a schematic diagram of the existing small patch fusion method considering the space competition division ability of adjacent patches: (a) the original skeleton line, (b) the adjusted skeleton line.

The fusion of small patches of land use data includes the simplification of spatial geometric features, the merging of semantic type levels and the adjacent relationship between adjacent plots. Ai Tinghua et al. (2002) proposed a better small patch fusion method based on Delaunay triangulation, that is, a weighted skeleton line segmentation method that takes into account the competitive segmentation ability of adjacent patch spaces. Its core steps are:

Step 1: Identify the merged patch candidate set that is smaller than the area threshold on the scale, such as patch a in Figure 1 (a);

Step 2: Calculate the spatial competitive segmentation ability under the constraints of spatial geometric features and semantic distance with adjacent patches b and c (assuming that the segmentation capabilities of adjacent patches b and c for small patches are 8 and 2, respectively) ;

Step 3: The division point is changed from the bisecting point of the original triangle side to the division point formed according to the division ability ratio. As shown in Figure 1(b), the small image spot a is divided and melted into the adjacent plots b, c.

Figure 2 is a schematic diagram of the shortcomings of the existing small patch melting methods considering local optimization: (a) the original land use data and the small patches (deepened part) in it, (b) based on Delaunay triangulation using local optimization The split line proposed by the algorithm (bold line), (c) the fusion result.

When using Delaunay triangulation to melt small patches of land use data, the most important thing is to determine the subdivision points of the patches. Existing small patch fusion methods rely more on directly using the local optimal calculation results of adjacent patches to subdivide and dissolve small patches (Podrenek, 2002; Van Smaalen, 2003). For example, based on the area of adjacent patches, sharing edge, local semantic distance, etc. (Ai Tinghua et al., 2010), neglecting the maintenance of the balance of the global land type area, resulting in a large change in the proportion of various land classification areas in the entire area before and after the melting operation (Li Jing et al. , 2014; Crown, 2009). Especially in the case where the land use data is relatively fragmented and there are many small patches, the local optimal method for patch fusion is insufficient in maintaining the balance of the global land type area:

Figure 2(a) shows the original land use data, and the small plots in the data are deepened, assuming that the rectangular frame is the entire study area; Figure 2(b) shows the local optimal algorithm based on the Delaunay triangulation ( Considering the area of adjacent patches, shared side length, and semantic distance), the corrected splitting line (bold line); the final small patch fusion result is shown in Figure 2(c).

The area ratio statistics of the first-level land type patches before and after the melting are carried out. The results are shown in Table 1. It can be seen that before and after the melting, the absolute value of the change in the area of the entire region is as high as 56%. big.

Table 1 Statistics on the proportion of first-level land-type map spots before and after melting

Note: This figure is only a schematic diagram, and the land use types in the area do not include all land types.

As shown in Figures 3a and 3b, the present application provides a method for spot melting that takes into account both local optimum and overall area balance, and is characterized in that, it includes:

S301, obtaining full-coverage vector image spot data of an image to be processed; and obtaining data information of adjacent spots in the image according to the data; wherein the data information of the adjacent spots includes: the area of adjacent spots, the shared edge Length and semantic distance; specifically, this application uses local optimal indicators and calculation: input a full-coverage vector map spot data, and propose three indicators of adjacent spot area, shared edge length, and semantic distance as the small graph spot melting process The calculation basis of the competitive ability of the local adjacent patch segmentation in the middle; specifically:

(1) Area of adjacent spots

The size of the adjacent patch area is the most intuitive factor that determines its ability to "occupy" a small patch. For the neighboring patches of a small patch, in principle, the patch with a larger area is more local than the patch with a smaller area. Excellent fusion object. The method of calculating the area of adjacent patches generally adopts the coordinate analysis method, and its mathematical model is as follows:

Wherein, the present invention uses the first-order adjacent field as the area constrained by the spatial pattern, i is the clockwise number of each node of the adjacent polygon, x _i is the abscissa of each node of the adjacent polygon, y _i+1 , y _i-1 are The ordinate of each node of the adjacent polygon.

(2) Shared edge length

The length of the shared edge is an important indicator for judging the spatial proximity between the patches. In landscape ecology, the larger the shared edge, the better the material and energy flow and transition between the patches. Therefore, the larger the shared edge The neighboring patches have better attribution of "competitiveness" to small patches. Shared edges are identified by adding semantic information to the topology. If the nodes of an arc segment have two different semantic information, the edge is a shared edge. The Euclidean distance method is generally used to calculate the distance (d) between the two nodes on the shared edge of the adjacent patch and the small patch, and its mathematical model is as follows:

(3) Semantic distance

Semantic proximity is the core element for judging the attribution of small patches (Liu et al., 2002). Van Oosterom (1995) proposed in the classic iterative merger algorithm that the semantic similarity between small-area patches and their neighboring patches is similar. The degree is used as the basis for judging the local optimum of adjacent patches. Based on the relevant theory of cartographic knowledge, the invention establishes a conditional semantic proximity model, and refines the semantic distance calculation method between land types. This method can not only prevent unreasonable conversion between land types (Yang Jun et al., 2013), but also ensure the merge operation of the same type of patches.

Land use data often care about the total amount of first- and second-level land categories, but in the original data, more refined third-level land categories are used as the classification and management unit of the patches. According to the classification and operation rules for merging data of geographic and national conditions in the "Census Contents and Indicators of Geographical National Conditions", the semantic proximity model of the present invention is created as follows:

Among them, X and Y _i are the two land types involved in the proximity calculation, _Yi is the land class of the merged source land class, X is the land class of the annexed target land class; m is the land class that has a semantic proximity relationship with X The number of categories, which specifies that the semantic distance between adjacent elements is 1 unit, and Distance(X,Y _i ) is the position of the merged source category in the semantically adjacent category set (first consider the category of the same parent category and then consider the proximity relationship with land types that are not of the same parent type).

S302, according to the data information of the adjacent spots, perform area pre-allocation on each of the small spots, to obtain the subdivision area of each adjacent spot to the small spots; wherein, the area smaller than the first specified threshold is A patch is used as a small patch; specifically, based on the CRITIC objective weighting method, a Split Ability Function (SAF) of adjacent patches is defined to calculate the spatial competitiveness of a small patch and its neighboring patches, and Delaunay is used to calculate the spatial competitiveness of adjacent patches. The triangulation network extracts the target skeleton line of the patch to pre-divide the area of the small patch.

Wherein, the S302 includes:

B1. Obtain the subdivision capability of each adjacent dot of each small dot for obtaining the subdivision area of the small dot;

B2. According to the ratio of the segmenting capability, obtain the segmented area of each adjacent patch of the small patch.

Wherein, the calculation model of the dissection capability described in step B1 is:

Among them, i=1, 2, 3 in S _i , S ₁ , S ₂ , and S ₃ represent three constraint indicators of adjacent patch area, shared edge length, and semantic distance, respectively, and _wi is the weight of each indicator; , a represents the small patch, and b _i represents the i-th adjacent patch. If either S1 or S2 is ₀ , then SAF is ₀ . SAF(a,b _i ) represents the ability of the i-th adjacent patch to obtain the segmented area of the small patch.

Wherein, in the step B2, according to the ratio of the division capability, the calculation model for obtaining the subdivision area of each adjacent spot to a small spot is:

Area _i =SAF(a, b _i )/SAF(a, b)*Area

Among them, Area _i is the subdivision area of the i-th adjacent dot to the small dot; SAF(a,b) is the sum of the subdivision capabilities of all adjacent dots; Area is the area of the small dot; SAF(a,b) _i ) represents the ability of the i-th adjacent patch to obtain the segmented area of the small patch.

S303: Count the areas of the first-level land types of each of the small patches after the area pre-allocation, and calculate the respective change rates of the areas of each type before and after the pre-allocation. specifically:

(1) Statistics on the area of each category

The area of the first-level land classes after spatial pre-classification is counted, and the change rate of the area of each class before and after the pre-classification is calculated. Given a threshold and a second specified threshold V, if the rate of change of at least one land type exceeds the threshold, record the difference U _i between all land types and the threshold, and iteratively adjust the area pre-segmentation results; if it does not exceed the threshold, directly The splitting line is determined and melted according to the area pre-segmentation result.

(2) Area balance iterative algorithm

The land types that need to be adjusted are divided into two categories, one is the land type whose area has increased beyond the threshold (denoted as A land type), and the other is the land type whose area reduction exceeds the threshold (denoted as B land type). The land type whose area change is within the threshold range after area pre-segmentation is recorded as land type C. The land types after area pre-division mainly include these three land types. Note that the total area of the small patch is N, and the main process of iterative adjustment of the land type area is as follows:

1) Find the patch information that needs to be adjusted

For the ground type that needs to be adjusted, first traverse all the patches adjacent to the patch of the ground type and the small patch, and record the information of the small patch and its adjacent patches;

2) Determine the effective adjustment patch information

Filter the adjacent patches of the small patch that only contain the patches of type A or type B, and obtain the effective adjustable patch. Note that the area of each effective adjustment patch is _Mi at this time, and the effective patch is The total area is M;

3) Statistically adjustable area of global scale land types

Assuming that the areas that need to be adjusted for the global land types A and B are U _a and U _b respectively, the maximum area that can be increased for land type C is U _c1 , and the maximum area that can be reduced is U _c2 ;

4) Count the scale of each small spot in the local area, and the area that can be adjusted by the land type

The global adjustment area obtained in step 3 is allocated to each valid small patch i, then its adjacent patches:

The area that needs to be adjusted for land type A is: N _ai =U _a /M*M _i

The area that needs to be adjusted for land type B is: N _bi =U _b /M*M _i

The maximum area that can be increased and adjusted for type C is: N _c1i =U _c1 /N*M _i

The maximum area that can be reduced and adjusted for type C is: N _c2i =U _c2 /N*M _i

5) Area adjustment

If there are only any two types of ground types A, B, and C in the adjacent patches of the small patch, then the area value adjusted by the two ground types is Min{N _ai , N _bi , N _ci };

If all three land types exist, the adjusted areas of land types A, B, and C are shown in the following table:

Table 4

Among them, when there are two or more A ground types (such as ground type A ₁ , ground type A ₂ ) in the adjacent patches, the area that needs to be adjusted in A ₁ is the area of local ground type A adjustment and the area of ground type A ₁ Adjust the product of the threshold proportions. Type B and Type C are treated the same.

6) Repeat iterative adjustment

Each small spot is adjusted according to step 5, and the adjusted area and the pre-divided area are added or subtracted correspondingly to obtain the area Area _i of the adjusted adjacent spots divided into small spots. Determine whether the adjusted area is within the threshold range through global area statistics, and iteratively adjust until all land types meet the conditions. The algorithm of the present invention sets the maximum number of iterations to 10,000 times, that is, the maximum number of iterations can be performed for 10,000 times.

S304, determine whether each rate of change is lower than the second specified threshold, if not, execute S305, and if so, execute S306.

S305, when there is a rate of change higher than the second specified threshold in each of the rates of change, record the difference between all land types and the second specified threshold, and iterate the area pre-segmentation result according to the area balance iterative algorithm Process until each of the rates of change is below the second specified threshold.

S306, when the respective change rates are all lower than the second specified threshold, determine the internal skeleton line of each of the small image spots, and split each small image spot into a plurality of fragments according to the skeleton lines; and The fragments are merged into adjacent patches so that the small patches are melted to form the merged full overlay patch data.

The S306 includes:

D1, according to the Delaunay triangulation method, determine the adjacent image spots that form a subdivision for the small image spots;

D2. Perform two-by-two calculations on the small image spots and the adjacent image spots, respectively, to obtain subdivision points; wherein, the calculation formula for obtaining the subdivision points is:

Among them, a represents the small patch, b represents a neighboring patch, c represents another neighboring patch, (x _b , y _b ), (x _c , y _c ), respectively represent the two ends of the edge in the triangular network Coordinates, where x _b and x _c represent the abscissa, and y _b and y _c represent the ordinate; Area(a,b), Area(a,c) are the divisions of adjacent patches b and c to small patch a, respectively area value.

D3. Generate a split line according to the split point, and split the small patch into a plurality of fragments accordingly, and merge the fragments into adjacent patches, so that the small patches are melted to The full-coverage patch data after the merger is formed.

Embodiment 2

Further, in order to better illustrate the technical solution of the present application, the applicant has also carried out the following experiments, relying on the WJ-III map workstation developed by the Chinese Academy of Surveying and Mapping Sciences, and embedding the local optimum proposed by the present invention based on the constrained Delaunay triangulation. The small graph spot melting method balanced with the overall area uses OpenMP to realize the small graph spot melting operation of geographic national conditions data in the C++ environment, and the rationality and effectiveness of the method of the present invention are verified. The experiment takes the full-coverage land use data of a county in Guangdong Province as an example. The scale of the original data is 1:10,000, and there are 2604 small patches that need to be melted. The types of land features are mainly natural features such as forest land, cultivated land, and water bodies, accounting for 49.6%, 26.2% and 7.0% respectively. Land features such as gardens, grasslands, houses and structures are scattered, and deserts and bare surfaces are scarce, only occupying than 0.01%. Among them, the simplified target scale is 1:100,000. The experimental environment is Microsoft Win7 64-bit operating system, the CPU is Intel Core I7-4790, a single machine with 8 cores and 8 threads, the main frequency is 3.2GHz, the memory is 16GB, and the solid-state drive is 1024GB.

According to the requirements of the technical regulations on the results of the census of geographical conditions and national conditions, if the scale of the map is greater than 1:500,000, the minimum area of the top map for each land type is shown in Table 2. The present invention uses this as the judgment standard for small map spots, and other scales are small maps. Refer to Table 2 for fine-tuning of the criteria for determining the spot.

Table 2 Minimum area of the upper image

(1) Superiority verification

The method of the present invention firstly uses the local optimal constraint to perform spatial pre-segmentation on the small spots in the test area, and counts the pre-segmented areas.

Table 3. Statistics on the area proportion of first-level land types before and after area pre-segmentation (unit: km ² )

Due to the lack of the standard for the change range of the area before and after the melting of the small patch, this experiment set the threshold of the change rate of the area of each type before and after the melting of the small patch to 5% according to the relevant literature (Liu Yaolin et al., 2009) and the data characteristics of the experimental area. It can be seen from Table 3 that only the local optimal method is used to dissolve small patches, and the area change rate of the land types of grassland and artificial structures exceeds the threshold of 5%. It shows that simply using the local optimal algorithm cannot guarantee the balance of the overall land type area. According to the algorithm of the present invention, iterative adjustment of the pre-divided area is required.

After 108 iterative adjustments in the experiment, it is ensured that the area change rate of all land types is less than the threshold after the small patches are melted. The present invention counts the adjustment values of the rate of change of each category at the 10th, 20th, 30th, 40th, 50th, 60th, 70th, 80th, 90th, 100th and 108th iterations of the algorithm, and adjusts the value of each category after each iteration adjustment. Statistical analysis was carried out on the area change rate of , and the results are shown in Figure 4 and Figure 5, respectively.

FIG. 4 is a schematic diagram showing the statistics of the adjustment value of the area change rate of each category according to the embodiment of the present invention.

FIG. 5 is a schematic diagram of iterative adjustment of the small patch melting area that takes into account the global optimum and is performed according to an embodiment of the present invention.

It can be seen from Figure 4 and Figure 5 that as the number of iterations increases, the area adjusted by each category gradually decreases. For each iterative adjustment, the adjustment values of grassland and artificial structures are all positive values, and the most negative values are cultivated land and forest land, followed by house construction and piled digging land, but due to the small reduction, it is not obvious in the figure . Each iteration of each category is a process of converging to the threshold. After the adjustment, the area change rate of all categories is less than the set threshold. It fully proves the feasibility of the method of the present invention and the superiority of ensuring the balance of the global land type area. Combining with the reliability verification of the method, it can be found that the method of the present invention can maintain the balance of the global land type area while ensuring the local space and semantic features, which proves the feasibility of the method.

(2) Reliability verification

In order to verify the reliability of the method of the present invention, that is, the melting results can still maintain the local optimum, the actual small spots were selected in the experimental area, and the adjusted melting results of the present invention were compared with the melting results preliminarily pre-divided according to the local optimum. Among them, grassland and structures whose area change rate exceeds the threshold are selected as test land types.

Respectively use x _i and y _i to represent the initial subdivision area of the grass or structure spot and the subdivided area after adjustment by the algorithm of the present invention, and calculate the difference (x _i -y _i ) between the two, that is, the adjustment area of each effective spot , where i represents the number of all the adjusted patches of the land type, and the statistical curve results of the adjusted area of each patch are shown in Figure 6 (due to the large number of adjusted patches, the present invention only selects the adjusted area in the top 100 statistics of the plot).

FIG. 6 is a schematic diagram showing the area statistics of the patch adjustment of grassland and artificial structures according to an embodiment of the present invention.

It can be seen from Figure 6 that the maximum adjusted area of grass patch is 283.3m ² , and the maximum adjusted area of structure patch is 112.6m ² , and the adjusted area has no significant change compared with the initial pre-division. In the adjustment patches of grass and structures, the first three patches (within the ellipse) with the largest adjustment area were selected respectively, and the adjusted melting results were compared with the initial subdivision results, as shown in Figure 6. The rectangular box A in Fig. 7 represents the result of dissolving the small patches of land use cover data by using the algorithm of the present invention, the rectangular boxes B1, B2 and B3 represent the three patches with larger grass adjustment areas, and the rectangular boxes C1, C2 and C3 represent the The artificial structures adjust the three patches with larger areas; the solid line shows the melting split line of the small patch using the algorithm of the present invention, and the dotted line represents the melting split line using only the locally optimal extraction of the patch.

FIG. 7 is a schematic diagram showing a result of dissolving a small patch of land use pavement data according to an embodiment of the present invention.

FIG. 8 is a schematic diagram showing the comparison of the melting area of the small map spot to the surrounding land type map spot according to the embodiment of the present invention.

The initial melting area of the small patches in the rectangular boxes of B1, B2, B3 and C1, C2, and C3 to the surrounding types of patches is compared with the melting area of each type of patches after adjustment. The results are shown in Figure 8. The dot-filled column represents the initial subdivision area of the small patch to the surrounding terrain based on the local optimum, and the blank-filled column represents the subdivision area adjusted according to the method of the present invention.

It can be seen from Figures 7 and 8 that, compared with the results obtained by considering only the local optimum, the fusion results of the small patches after iterative adjustment using the method of the present invention, no matter from the perspective of spatial features or specific area differences, the surrounding patches are relatively small. The segmentation capabilities of the patches are not much different, and the two results maintain a high consistency in the local features, which fully proves that the method of the present invention can still maintain the reliability of the local spatial proximity and semantic distance features.

To sum up, the patch melting method proposed in the present application can maintain the balance of the overall land type area before and after melting while taking into account the optimal local spatial pattern of the patch during the melting process. It is beneficial to better maintain the consistency of the land types before and after the merger of the map patches.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the scope of the present invention. within the scope of protection.

Claims (9)

1. A small image spot melting method considering local optimization and integral area balance is characterized by comprising the following steps:

A. acquiring full-coverage vector pattern spot data of an image to be processed; acquiring data information of adjacent pattern spots according to the data; wherein, the data information of the adjacent pattern spots at least comprises one of the following data information: the area of the adjacent pattern spot, the length of the shared edge and the semantic distance;

B. according to the data information of the adjacent image spots, carrying out area pre-allocation on the small image spots in the image to obtain the subdivision area of each adjacent image spot on the small image spot; wherein, the image spot with the area smaller than the first designated threshold value is taken as a small image spot;

C. counting the area of the first-level land class of each small map spot after area pre-allocation, and calculating each change rate of each land class area before and after the pre-allocation;

D. when the change rates are lower than a second specified threshold value, determining an internal skeleton line of each small image spot, and respectively splitting each small image spot into a plurality of fragments according to the skeleton lines; and merging the patches into patches adjacent thereto so that the small patches are thawed to form merged full-coverage patch data.

2. The method of claim 1, wherein step C further comprises:

and when the change rate of each change rate is higher than a second specified threshold value, recording the difference value between all the land categories and the second specified threshold value, and performing iterative processing on the area pre-classification result according to an area balance iterative algorithm until each change rate is lower than the second specified threshold value.

3. The method of claim 1, wherein the step a of obtaining the computational model of the area of the adjacent patches is:

wherein i is the clockwise number of each node of the adjacent polygon, xi is the abscissa of each node of the adjacent polygon, and yi +1 and yi-1 are the ordinates of each node of the adjacent polygon.

4. The method according to claim 1, wherein the step a of obtaining the calculation model of the shared edge length is:

wherein xi and yi are respectively the abscissa and the ordinate of an end point of the shared edge; xi +1 and yi +1 represent the abscissa and ordinate, respectively, of the other end point of the shared edge.

5. The method of claim 1, wherein step B comprises:

b1, obtaining the subdivision capability of each adjacent image spot of each small image spot on the subdivision area for obtaining the small image spot;

and B2, acquiring the subdivision area of each adjacent image spot to the small image spot according to the proportion of the subdivision capacity.

6. The method of claim 5, wherein the subdivision capability calculation model of step B1 is:

wherein, i in Si is 1,2,3, S1, S2, S3 respectively represent three constraint indexes of adjacent patch area, shared edge length, semantic distance, and wi is the weight of each index; where a represents the small spot and bi represents the ith neighboring spot.

7. The method according to claim 6, wherein the step B2 is to obtain a calculation model of the subdivision area of each adjacent plaque to the small plaque according to the ratio of the subdivision ability, wherein the calculation model comprises:

Areai＝SAF(a，bi)/SAF(a，b)*Area

wherein, the area is the subdivision area of the ith adjacent image spot to the small image spot; SAF (a, b) is the sum of all adjacent plaque dissection capabilities; area is the Area of the small pattern spot; SAF (a, bi) shows the subdivision capability of the ith adjacent pattern spot for acquiring the subdivision area of the small pattern spot.

8. The method of claim 1, wherein step D comprises:

d1, determining adjacent image spots for forming subdivision on the small image spots according to a Delaunay triangulation method;

d2, respectively carrying out pairwise calculation on the small pattern spots and the adjacent pattern spots to obtain split points;

d3, generating a splitting line according to the splitting points, splitting the small image spots into a plurality of fragments according to the splitting line, merging the fragments with adjacent image spots so that the small image spots are fused to form merged full-coverage image spot data.

9. The method of claim 8, wherein the calculation formula for obtaining the split point is:

wherein a represents the small patch, b represents a neighboring patch, c represents another neighboring patch, (xb, yb), (xc, yc) respectively represent coordinates of two end points of the side in the triangulation network, wherein xb and xc represent abscissa, yb and yc represent ordinate; area (a, b) and Area (a, c) are values of the division areas of the small spots a by the adjacent spots b and c, respectively.

Images