A method for evaluating geometric accuracy of vector map data

Abstract

An embodiment of the present invention provides a method for evaluating the geometric accuracy of vector map data, which relates to the field of geographic information data processing technology. The method includes the following steps: obtaining digital orthophoto data that is consistent with the regional location of the vector map data to be evaluated, generating reference vector data; using entity elements in the vector map data to be evaluated as entity elements to be evaluated, using entity elements in the reference vector data as reference entity elements, performing similarity matching between the entity elements to be evaluated and the reference entity elements, and generating a mapping relationship for entities with the same name based on the matching results; calculating the mean square error of the distance between the centroid coordinate points of each group of entities with the same name in a sample data set of entities with the same name, and using the mean square error as the geometric accuracy of the vector map data to be evaluated. This solution, through automated geometric accuracy evaluation, can reduce the workload of manual evaluation during batch operations, fully utilize computing resources, and save costs.

Description

Technical Field

[0001] The present invention relates to the technical field of geographic information data processing, and in particular to a method for evaluating the geometric accuracy of vector map data. Background Art

[0002] With the development of globalization, the application areas of vector map data are constantly expanding, and the requirements for geometric accuracy are becoming increasingly higher. In actual production and updating, errors will occur in the collection, processing, storage, and use of vector map data. For example, manual measurement may cause errors due to improper operation, insufficient accuracy of cartographic data, and other factors; the conversion between geographic coordinate systems and projected coordinate systems may also introduce errors, especially when converting over large areas; when processing large-scale vector data, cropping and splicing operations are required, which may introduce errors due to loss of accuracy or improper boundary processing; when converting between different data formats, errors may also be caused by format differences. Therefore, when using vector map data for certain important analysis and decision-making applications, it is necessary to evaluate the geometric accuracy of the data to reduce the impact of geometric errors on decision-making results.

[0003] Evaluating the geometric accuracy of vector map data is a data analysis process. Traditionally, this approach involves manually overlaying vector map data with DOM (digital orthophoto) data. Measurement tools are then used to calculate the offset distance between features in the DOM (digital orthophoto) and the same features in the vector data. The geometric accuracy of the data is then described by measuring the mean square error between the distances between multiple features with the same name. This process is complex and can lead to inaccurate evaluation results due to manual errors.

[0004] Currently, there are two main problems in the geometric accuracy assessment of vector map data:

[0005] First, there is a lack of automated geometric accuracy assessment tools. Current assessment tools rely on manual interaction, resulting in low assessment efficiency and long assessment time.

[0006] Secondly, the manual evaluation method has a low accuracy rate. During the manual evaluation process, the offset distance between the reference element obtained by collecting DOM and the homonymous point of the element to be evaluated is calculated. If the homonymous point is not selected accurately, the evaluation result will be inaccurate. Summary of the Invention

[0007] In view of this, an embodiment of the present invention provides a method for evaluating the geometric accuracy of vector map data to solve the technical problems of non-automatic and low accuracy in the prior art evaluation of the geometric accuracy of vector map data. The method includes:

[0008] Obtain digital orthophoto data that is consistent with the regional location of the vector map data to be evaluated, convert the digital orthophoto data into geographic vector data, simplify and perform consistency processing on the geographic vector data to generate reference vector data, wherein the reference vector data is consistent with the projection coordinate system and scale of the vector map data to be evaluated;

[0009] Taking the entity element in the vector map data to be evaluated as the entity element to be evaluated, taking the entity element in the reference vector data as the reference entity element, obtaining the coordinate point of the centroid of the entity element to be evaluated, obtaining the coordinate point of the reference centroid of the reference entity element, matching the entity element to be evaluated with the reference entity element with the centroid of the entity element to be evaluated and the reference centroid of the entity element as the coincidence point, and generating a mapping relationship of entities with the same name according to the matching result;

[0010] The mapping relationships of entities with the same name are evenly distributed throughout the entire area to generate a sample dataset of entities with the same name. The mean square error of the distance between the centroid coordinate points of each group of entities with the same name in the sample dataset is calculated, and the mean square error is used as the geometric accuracy of the vector map data to be evaluated.

[0011] Compared with the prior art, the at least one technical solution adopted in the embodiments of this specification can achieve the following beneficial effects:

[0012] Through automated geometric accuracy assessment, the workload of manual assessment can be reduced during batch operations, computing resources can be fully utilized, and costs can be saved. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0014] Figure 1 This is a flow chart of a method for evaluating geometric accuracy of vector map data provided by an embodiment of the present invention;

[0015] Figure 2 is a flowchart of a method for implementing the geometric accuracy assessment of vector map data provided by an embodiment of the present invention;

[0016] Figure 3 Schematic diagram of a deep convolutional network structure of a method for geometric accuracy assessment of vector map data provided by an embodiment of the present invention;

[0017] Figure 4 It is a binarized raster image thumbnail of a method for evaluating geometric accuracy of vector map data provided by an embodiment of the present invention;

[0018] Figure 5 A geographic vector data diagram of a method for evaluating geometric accuracy of vector map data provided by an embodiment of the present invention;

[0019] Figure 6 Schematic diagram of vector data before and after simplification processing in a method for evaluating geometric accuracy of vector map data provided by an embodiment of the present invention;

[0020] Figure 7 3. It is a schematic diagram of centroid calculation of a method for geometric accuracy assessment of vector map data provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0021] The embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0022] The following describes the embodiments of the present application through specific examples, and those skilled in the art can easily understand other advantages and effects of the present application from the contents disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The present application can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that, in the absence of conflict, the features in the following embodiments and embodiments can be combined with each other. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without making creative work are within the scope of protection of this application.

[0023] Geometric accuracy assessment is an important part of vector map data selection and application. It is not only related to the accuracy and reliability of vector map data, but also directly affects the quality of basic geographic information data.

[0024] An embodiment of the present invention proposes a method for geometric accuracy assessment of vector map data. The main ideas of geometric accuracy assessment are as follows: First, a deep convolutional network is used to automatically extract building features from digital orthophotos to obtain binary raster image data containing building features. The data obtained by vectorizing the binary raster image data is used as the reference vector data for geometric accuracy assessment; second, feature similarity matching is used to establish a mapping relationship between the vector map data to be evaluated and the entities with the same name in the reference vector data; finally, the mean square error of the distance between the centroid coordinate points of the entities with the same name is calculated, and the mean square error is used as the indicator for geometric accuracy assessment.

[0025] This method for assessing the geometric accuracy of vector map data is technologically advanced and highly automated. Compared to traditional manual methods, it not only improves the efficiency of vector map data assessment but also avoids the inaccuracies often associated with manual evaluation. Through scientific and quantitative geometric accuracy assessment, it provides a basis for data selection and application.

[0026] In an embodiment of the present invention, a method for evaluating geometric accuracy of vector map data is provided. Figure 1 and Figure 2 As shown, the method includes:

[0027] Step S101: obtaining digital orthophoto data that is consistent with the regional location of the vector map data to be evaluated, converting the digital orthophoto data into geographic vector data, simplifying and performing consistency processing on the geographic vector data to generate reference vector data, wherein the reference vector data is consistent with the projection coordinate system and scale of the vector map data to be evaluated;

[0028] Step S102: Using the entity element in the vector map data to be evaluated as the entity element to be evaluated, and using the entity element in the reference vector data as the reference entity element, obtaining the coordinate point of the centroid of the entity element to be evaluated, obtaining the coordinate point of the reference centroid of the reference entity element, matching the entity element to be evaluated with the reference entity element with the centroid of the entity element to be evaluated and the reference centroid of the entity element as the coincidence point, and generating a mapping relationship of entities with the same name based on the matching result;

[0029] Step S103: evenly distribute the mapping relationships of entities with the same name to the entire area, generate a sample dataset of entities with the same name, calculate the mean square error of the distance between the centroid coordinate points of each group of entities with the same name in the sample dataset, and use the mean square error as the geometric accuracy of the vector map data to be evaluated.

[0030] In specific implementation, in order to make geographic vector data more consistent with the surface building elements of digital orthophotos and to achieve the goal of making the accuracy of geographic vector data more consistent with the requirements of benchmark vector data, the following steps are performed to obtain digital orthophoto data that is consistent with the regional location of the vector map data to be evaluated, convert the digital orthophoto data into geographic vector data, simplify and process the geographic vector data for consistency, and generate benchmark vector data:

[0031] A deep convolutional network is trained to extract binary data of building elements from digital orthophoto data and generate binary raster image data. The binary raster image data is vectorized to generate geographic vector data, wherein the geographic vector data has the same spatial range as the digital orthophoto data. The geographic vector data is consistency processed to make the projection coordinate system and scale of the vector map data to be evaluated consistent with those of the geographic vector data, thereby generating consistency-processed vector data, which is used as the baseline vector data.

[0032] In specific implementation, in order to use the deep convolutional network to extract building element binary images from DOM data (digital orthophoto data), the deep convolutional network is trained through the following steps:

[0033] Set the bias parameter b of the convolution layer of the deep convolutional network x and weight parameter W x , repeat the following steps until the deep convolutional network training is completed: input the raster image in the training set to the convolution kernel f of the convolution layer x In the example, the bias parameter b x Get the convolution layer C x ; After taking the average of the four pixels in each adjacent area of ​​the raster image, a merged pixel is generated, and the merged pixel is averaged by the weight parameter W by activating the Sigmoid function. x+n and bias parameter b x+n Generate mapping image S x+n , where n is the number of training times,

[0034] Specifically, the deep convolutional network is trained multiple times.

[0035] Specifically, collect digital orthophoto data that is consistent with the area of ​​the vector map data to be evaluated, and ensure that the resolution of the digital orthophoto data is not less than 0.5 meters.

[0036] Specifically, the RCNN network is used to extract binary images of building elements from DOM data (digital orthophoto data), and high-precision binary raster image data of surface building elements is obtained through georeferencing.

[0037] RCNN is a deep convolutional network that can extract building features through sample training. Its convolution process uses a trainable convolution kernel f x Perform convolution operation on the input DOM data (digital orthophoto data) by adding bias b x Get the convolution layer C x . The average of every 4 pixels in the DOM data (digital orthophoto data) is merged into one pixel, and the weighted W is first x+1Add bias b x+1 , using the activation Sigmoid function to form a mapping image S that is reduced to 1 / 4 x+1 The subsampling process is used. Different convolution kernels are used to extract difference features in the convolution process, and the target contour recognition is optimized in the process of weight sharing. The secondary features are extracted through the downsampling process, which reduces the dimension of DOM data (digital orthophoto data) and increases the robustness of target contour recognition. The above process is repeated many times to obtain a deep convolutional network. The structure of the deep convolutional network is as follows: Figure 3 shown.

[0038] Use RCNN deep convolutional network for sample training to generate a building feature extraction algorithm model, input DOM data (digital orthophoto data) into the algorithm model to obtain the inferred building feature binary image. Automatically align the binary image with the original DOM data (digital orthophoto data) to obtain binary raster image data with geographic coordinates (such as Figure 4 shown).

[0039] In specific implementation, the following steps are used to simplify geographic vector data and filter out extremely small surface features generated by noise to retain valid data:

[0040] Before the geographic vector data is processed for consistency, it is simplified and filtered to remove aliasing, jitter and invalid data.

[0041] Specifically, the binary raster image data is vectorized using the internal point diffusion method to generate vector data consistent with the DOM space range (such as Figure 5 Secondly, due to the binary raster image data pixels, the generated vector data boundaries have jagged and jitter phenomena, and the vector data needs to be simplified (such as Figure 6 (as shown in the figure), and also filter out extremely small polygon features generated by noise in the data to retain valid data. Finally, comprehensive vector data processing is performed to ensure that the projection coordinate system and scale of the vector map data to be evaluated are consistent with those of the baseline vector data.

[0042] In specific implementation, the following steps are performed to obtain the coordinate point of the centroid of the entity element to be evaluated:

[0043] Get the coordinates of all vertices in the entity element to be evaluated and generate a vertex coordinate set; calculate the coordinate point of the centroid to be evaluated (x c ,y c ),in, (x i ,y i) is the coordinate of the i-th vertex in the vertex coordinate set, and n is the total number of vertices.

[0044] Specifically, such as Figure 7 As shown, the centroid coordinates are the average of the coordinates of all points inside the polygon. If there are n points on the edge of the polygon, their coordinates are (x1, y1), (x2, y2), ..., (x n ,y n ), then the centroid coordinates (x c ,y c ) is calculated as follows:

[0045] In specific implementation, in order to use the centroid points of two entity elements with the same name as the coincidence point, perform similarity matching, and establish a mapping relationship between entities with the same name, the following steps are used to match the entity element to be evaluated with the reference entity element using the centroid coordinate point to be evaluated and the reference centroid coordinate point as the coincidence point, and generate a mapping relationship between entities with the same name based on the matching results:

[0046] Set the area range for restricted matching, and calculate the centroid distance between the centroid coordinate point to be evaluated and the reference centroid coordinate point within the area range for restricted matching; set a predetermined threshold range, and if the centroid distance is less than the predetermined threshold range, the entity features to be evaluated that meet the conditions will form a same-name entity mapping relationship with the reference entity features; if the same-name entity mapping relationship formed by the entity features to be evaluated and the reference entity features is not one-to-one corresponding, calculate the geometric intersection and geometric union of the same-name entities in the same-name entity mapping relationship, perform secondary matching based on the geometric intersection and geometric union, and generate the same-name entity mapping relationship based on the result of the secondary matching.

[0047] In specific implementation, the following steps are used to calculate the geometric intersection and geometric union of entities with the same name in the mapping relationship of entities with the same name, perform secondary matching based on the geometric intersection and geometric union, and generate the mapping relationship of entities with the same name based on the result of the secondary matching:

[0048] Set a fixed spatial distance, take the centroid coordinate point to be evaluated as the center, and obtain the geometric intersection of the entity feature to be evaluated and the reference entity feature within the fixed spatial distance; take the centroid coordinate point to be evaluated as the center, and obtain the geometric union of the entity feature to be evaluated and the reference entity feature within the fixed spatial distance; calculate the ratio of the geometric intersection and the geometric union, match the entity feature to be evaluated with the reference entity feature whose ratio is closest to 1, and generate a mapping relationship between entities with the same name.

[0049] Specifically, the method uses the centroid coordinates of the entity elements extracted from the vector map data to be evaluated and the DOM data (digital orthophoto data) as the center, and performs entity element similarity matching within a fixed spatial distance. By obtaining the intersection and union information of the entity elements and calculating the ratio of the intersection information to the union information, the similarity of the entity elements is determined and a mapping relationship between entities with the same name is established.

[0050] Specifically, to obtain the most similar entity data with the same name, certain rules need to be set. For example, if there are multiple entities similar to the vector data entity element, the matching area must be limited first. The relationship between entities with the same name is determined by ensuring that the centroid coordinates of the baseline vector data entity element and the vector map data entity element to be evaluated are within a certain predetermined threshold range. Secondly, if there are multiple similar entities with the same name within a certain range, the intersection and union information ratio method is used to determine the relationship between entities with the same name. The entity element to be evaluated that is closest to 1 is taken as the entity with the same name, and the mapping relationship between entities with the same name is constructed.

[0051] Specifically, the geometric union is the part generated by merging the spatial features of two different layers in spatial position. The geometric intersection is the part where the spatial features of two different layers intersect in spatial position.

[0052] In the specific implementation, in order to calculate the mean error of the actual coordinate distance of the centroid points of multiple entities with the same name, the mean error is used as the geometric accuracy evaluation result. The mapping relationship of entities with the same name is evenly distributed throughout the entire area through the following steps to generate a sample dataset of entities with the same name:

[0053] A segmentation grid is set, and the entire area where the vector map data to be evaluated is divided into multiple sub-areas according to the segmentation grid; in each sub-area, a pair of entity mapping relationships with the same name is randomly selected until all sub-areas are selected to generate a sample dataset of entities with the same name.

[0054] Specifically, after obtaining multiple mapping relationships for entities with the same name, entities with the same name are sampled evenly across the entire region to ensure uniform distribution of the entity samples. During sampling, the entire region can be divided into 3×3, 4×4, or 5×5 grids. Within each subregion, a pair of entity mapping relationships with the same name is randomly sampled to establish a sample dataset of entities with the same name.

[0055] In specific implementation, the following steps are used to calculate the mean square error of the distance between the centroid coordinate points of each group of entities with the same name in the sample dataset of entities with the same name:

[0056] Calculate the distance between the centroid coordinate points of each group of entities with the same name in the sample dataset of entities with the same name; based on the distance between the centroid coordinate points of each group of entities with the same name, calculate the average distance between the centroid coordinate points of all entities with the same name in, n is the number of entities with the same name; by the average Calculate the mean error Among them, x i is the distance between the centroid coordinate points of the i-th group in the sample dataset of entities with the same name, and n is the number of calculations.

[0057] The embodiments of the present invention achieve the following technical effects:

[0058] The vector map data geometric accuracy assessment method of the embodiment of the present invention combines the feature extraction capabilities of deep learning algorithms and traditional vector data processing methods, giving full play to their respective advantages and improving the accuracy and reliability of data geometric assessment; through deep learning methods, it can realize automatic processing of large amounts of data, improving processing efficiency and work efficiency; through automated geometric accuracy assessment, during batch operations, it can reduce the workload of manual assessment, fully utilize computing resources, and save costs.

[0059] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A method for evaluating geometric accuracy of vector map data, characterized in that: include: Acquiring digital orthophoto data that is consistent with the regional location of the vector map data to be evaluated, converting the digital orthophoto data into geographic vector data, simplifying and performing consistency processing on the geographic vector data to generate reference vector data, wherein the reference vector data is consistent with the projection coordinate system and scale of the vector map data to be evaluated; The method includes: taking the entity element in the vector map data to be evaluated as the entity element to be evaluated, taking the entity element in the reference vector data as the reference entity element, obtaining the centroid coordinate point to be evaluated of the entity element to be evaluated, obtaining the reference centroid coordinate point of the reference entity element, performing similarity matching between the entity element to be evaluated and the reference entity element with the centroid coordinate point to be evaluated and the reference centroid coordinate point as coincidence points, and generating a mapping relationship of entities with the same name according to the matching results, including: Set a restricted matching area range, calculate the centroid distance between the centroid coordinate point to be evaluated and the reference centroid coordinate point within the restricted matching area range; set a predetermined threshold range, and if the centroid distance is less than the predetermined threshold range, form a same-name entity mapping relationship between the entity element to be evaluated and the reference entity element that meets the conditions; if the same-name entity mapping relationship formed by the entity element to be evaluated and the reference entity element is not one-to-one corresponding, calculate the geometric intersection and geometric union of the same-name entities in the same-name entity mapping relationship, perform secondary matching based on the geometric intersection and the geometric union, and generate a same-name entity mapping relationship based on the result of the secondary matching; Calculating the geometric intersection and the geometric union of the entities with the same name in the entity mapping relationship with the same name, performing secondary matching based on the geometric intersection and the geometric union, and generating the entity mapping relationship with the same name based on the result of the secondary matching, including: Setting a fixed spatial distance, taking the centroid coordinate point to be evaluated as the center, and obtaining the geometric intersection of the entity element to be evaluated and the reference entity element within the fixed spatial distance; taking the centroid coordinate point to be evaluated as the center, and obtaining the geometric union of the entity element to be evaluated and the reference entity element within the fixed spatial distance; calculating the ratio of the geometric intersection to the geometric union, matching the entity element to be evaluated with the reference entity element whose ratio is closest to 1, and generating a mapping relationship of entities with the same name; Evenly distribute the mapping relationships of entities with the same name to the entire area to generate a sample dataset of entities with the same name, calculate the mean square error of the distance between the centroid coordinate points of each group of entities with the same name in the sample dataset of entities with the same name, and use the mean square error as the geometric accuracy of the vector map data to be evaluated; Evenly distribute the mapping relationships of entities with the same name to the entire region to generate a sample dataset of entities with the same name, including: A segmentation grid is set, and the entire area where the vector map data to be evaluated is divided into multiple sub-areas according to the segmentation grid; in each of the sub-areas, a pair of entity mapping relationships with the same name is randomly selected until all the sub-areas are selected, thereby generating a sample dataset of entities with the same name.

2. The method for geometric accuracy assessment of vector map data according to claim 1, wherein: Obtaining the coordinate point of the centroid of the entity element to be evaluated, including: Obtaining the coordinates of all vertices in the entity element to be evaluated and generating a vertex coordinate set; Through the vertex coordinate set, the centroid coordinate point to be evaluated is calculated ( ),in, ， ， ( ) is the first vertex coordinate set The coordinates of the vertices, is the total number of vertices.

3. The method for evaluating geometric accuracy of vector map data according to claim 1, wherein: Calculating the mean square error of the distance between the centroid coordinate points of each group of entities with the same name in the entity sample data set, including: Calculating the distance between the centroid coordinate points of each group of entities with the same name in the entity sample data set; According to the distance between the centroid coordinate points of each group of entities with the same name, the average distance between the centroid coordinate points of all entities with the same name is calculated. ,in, , is the number of entities with the same name; By the average Calculate the mean error ,in, It is the first The distance between the centroid coordinates of the group, is the number of calculations.

4. The method for evaluating geometric accuracy of vector map data according to claim 3, wherein: Calculating the distance between the centroid coordinate points of each group of entities with the same name in the entity sample data set with the same name includes: Respectively obtaining the centroid coordinate point of each entity with the same name in each group of entities with the same name, to obtain two centroid coordinate points; Calculate the distance between the two centroid coordinate points.

5. The method for geometric accuracy assessment of vector map data according to any one of claims 1 to 4, characterized in that: Acquiring digital orthophoto data that is consistent with the regional location of the vector map data to be evaluated, converting the digital orthophoto data into geographic vector data, simplifying and performing consistency processing on the geographic vector data to generate reference vector data, including: Training a deep convolutional network, extracting binary data of building elements in the digital orthophoto data through the deep convolutional network, and generating binary raster image data; Vectorizing the binarized raster image data to generate geographic vector data, wherein the geographic vector data is consistent with the spatial range of the digital orthophoto data; The geographic vector data is subjected to consistency processing to make the projection coordinate system and scale of the vector map data to be evaluated consistent with those of the geographic vector data, to generate consistent processed vector data, and the consistent processed vector data is used as the reference vector data.

6. The method for evaluating geometric accuracy of vector map data according to claim 5, wherein: Also includes: Before the geographic vector data is subjected to consistency processing, the geographic vector data is subjected to simplification processing and filtering processing to remove aliasing, jitter and invalid data.

7. The method for evaluating geometric accuracy of vector map data according to claim 5, wherein: Training deep convolutional networks, including: Set the bias parameters of the convolutional layer of the deep convolutional network and weight parameters , repeat the following steps until the deep convolutional network training is completed: Input the raster image in the training set to the convolution kernel of the convolution layer In the example, the bias parameter Get the convolutional layer ; After taking the average of the four pixels in each adjacent area of ​​the raster image, a merged pixel is generated and activated. Sigmoid Function, the merged pixels are passed through the weight parameter and the bias parameters Generate a mapping image ,in, is the number of training sessions.

Images