POI data correction method and device based on image segmentation, equipment and storage medium

Abstract

The invention discloses a POI data correction method, device and equipment based on image segmentation and a storage medium, and the method comprises the steps: constructing a training image data set which comprises A-source road network image data and B-source road network image data; based on the A-source road network image data and the B-source road network image data, performing iterative training on a preset POI correction matching model, and when a preset convergence condition is met, obtaining a trained POI correction matching model; inputting to-be-detected data into the trained preset POI correction matching model to obtain a POI point matching result and a surrounding road network similarity comparison result; and on the basis of the POI point matching result and the similarity comparison result of the surrounding road network, updating the POI points in the road network to be updated. According to the method, the image data set is constructed based on the multi-source road network data, model training is carried out, the POI point surrounding road network similarity and the POI point matching result can be obtained at the same time based on the trained model, the POI point matching accuracy can be improved, and the map data updating efficiency can be improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a POI data correction method, device, computer equipment and storage medium based on image segmentation. Background Art

[0002] Point of Information (POI) is a common term in the field of geographic information. A POI can represent a house, a shop, or a bus or subway station, etc. POI describes the location of an information point through name, geographic coordinates, location category and other information. In map navigation applications, users can query the location information of POI points through various information.

[0003] The amount of POI data in map data is huge, and the POI information in the real world is often updated and changed. In order to provide users with accurate and timely POI data in navigation applications, POI data needs to be updated accurately and efficiently. Therefore, the updating method of POI data has important practical application significance.

[0004] Existing POI correction and update methods are usually based on regional threshold calculation, establishment of irregular triangulated networks and other methods to compare and update new POI point data with the original POI data within a certain area. Relying on these traditional correction methods is prone to misjudgment and requires a lot of manual work for verification and correction. The accuracy of POI correction needs to be further improved. Summary of the invention

[0005] Based on this, it is necessary to provide a POI data correction method, device, computer equipment and storage medium based on image segmentation to address the above technical problems, so as to solve at least one problem in the above-mentioned prior art.

[0006] In a first aspect, the embodiment of the present application is implemented as follows: a POI data correction method based on image segmentation is provided, comprising the following steps:

[0007] Constructing a training image data set, wherein the training image data set includes A-source road network image data and B-source road network image data;

[0008] Based on the A-source road network image data and the B-source road network image data, iteratively train the preset POI deviation correction matching model, and when the preset convergence condition is met, obtain a trained POI deviation correction matching model;

[0009] Input the data to be detected into the trained preset POI deviation correction matching model to obtain the POI point matching result and the surrounding road network similarity comparison result;

[0010] Based on the POI point matching result and the surrounding road network similarity comparison result, the POI points in the road network to be updated are updated.

[0011] In one embodiment, the data to be detected includes first A-source road network image data within M meters around the POI point to be matched in the A-source road network, second A-source road network image data within N meters around the POI point to be matched, and first B-source road network image data corresponding to the first A-source road network image data, M>N.

[0012] In one embodiment, the data to be detected is input into the trained preset POI deviation correction matching model to obtain the POI point matching result and the surrounding road network similarity comparison result, including:

[0013] Superimposing the first A-source road network image data and the second A-source road network image data, and extracting features from the superimposed image data to obtain an A-source road network feature vector;

[0014] Based on the B-source road network image data and the second A-source road network image data, a sampling feature vector and a splicing feature vector are obtained;

[0015] Perform feature decoding and pixel point classification on the A-source road network feature vector and the spliced ​​feature vector to obtain the POI point matching result;

[0016] Based on the POI point matching result and the sampling feature vector, the similarity between the peripheral road network of the A source road network and the peripheral road network of the B source road network is calculated to obtain the peripheral road network similarity comparison result.

[0017] In one embodiment, obtaining a sampling feature vector and a splicing feature vector based on the B-source road network image data and the second A-source road network image data includes:

[0018] Extracting image features from the B-source road network image data to obtain a B-source road network feature vector;

[0019] Up-sampling / down-sampling the second A-source road network image data to obtain a sampling feature vector;

[0020] The B-source road network feature vector is concatenated with the sampling feature vector to obtain a concatenated feature vector.

[0021] In one embodiment, the updating of the POI points in the road network to be updated based on the POI point matching result and the surrounding road network similarity comparison result includes:

[0022] Based on the surrounding road network similarity comparison result, when the surrounding road network similarity is greater than / equal to a preset similarity threshold, the POI matching result is subjected to an inverse transformation of longitude and latitude to obtain the corresponding POI point longitude and latitude information;

[0023] Update the corresponding POI point position in the road network to be updated based on the latitude and longitude information of the POI point; or

[0024] When the similarity of the surrounding road network is less than a preset similarity threshold, manual verification is performed to update the position of the POI point in the road network to be updated based on the verification result.

[0025] In one embodiment, constructing a training image dataset includes:

[0026] Collect A-source road network training data, and extract first A-source road network training data within M meters around each POI point and second A-source road network training data within N meters around each POI point from the A-source road network training data, with M>N;

[0027] Converting the first A-source road network training data into first A-source road network image training data, and converting the first A-source road network training data into second A-source road network image training data;

[0028] Collecting B-source road network training data, and extracting first B-source road network training data corresponding to the first A-source road network training data and second B-source road network training data corresponding to the second A-source road network training data from the open source road network training data;

[0029] Converting the first B-source road network training data into first B-source road network image training data, and converting the second B-source road network training data into second B-source road network image training data;

[0030] The first A-source road network image training data, the second A-source road network image training data, the first B-source road network image training data, and the second B-source road network image training data are used as image pairs corresponding to POI points;

[0031] The training image data set is constructed based on the image pairs of each of the POI points.

[0032] In one embodiment, the method further includes:

[0033] Performing a similarity comparison on the A-source road network image data and the B-source road network image data to obtain a comparison similarity;

[0034] When the comparison similarity is greater than / equal to a preset threshold, the position of the POI point to be matched, the position of the road network around the POI point to be matched, and the similar road category are marked in the B-source road network image data;

[0035] When the comparison similarity is less than a preset threshold, the matchable road network positions around the POI point to be matched and different road categories are marked in the B-source road network image data.

[0036] In a second aspect, a POI data correction device based on image segmentation is provided, comprising:

[0037] A training image data set construction unit, used to construct a training image data set, wherein the training image data set includes A-source road network image data and B-source road network image data;

[0038] A model training unit, based on the A-source road network image data and the B-source road network image data, iteratively trains the preset POI deviation correction matching model, and obtains a trained POI deviation correction matching model when a preset convergence condition is met;

[0039] A prediction unit, used for inputting the data to be detected into the trained preset POI deviation correction matching model to obtain the POI point matching result and the surrounding road network similarity comparison result;

[0040] The updating unit is used to update the POI points in the road network to be updated based on the POI point matching result and the surrounding road network similarity comparison result.

[0041] In a third aspect, a computer device is provided, comprising a memory, a processor, and computer-readable instructions stored in the memory and running on the processor, wherein when the processor executes the computer-readable instructions, the POI data correction method based on image segmentation as described above is implemented.

[0042] In a fourth aspect, a readable storage medium is provided, on which computer-readable instructions are stored. When the computer-readable instructions are executed by a processor, the POI data correction method based on image segmentation as described above is implemented.

[0043] The above-mentioned POI data correction method, device, computer equipment and storage medium based on image segmentation, the method implementation includes: constructing a training image data set, the training image data set includes A source road network image data and B source road network image data; based on the A source road network image data and the B source road network image data, iteratively training the preset POI correction matching model, and when the preset convergence condition is met, obtaining the trained POI correction matching model; inputting the data to be detected into the trained preset POI correction matching model to obtain the POI point matching result and the surrounding road network similarity comparison result; based on the POI point matching result and the surrounding road network similarity comparison result, updating the POI points in the road network to be updated. In the embodiment of the present application, an image data set is constructed based on multi-source road network data, and the relative geometric position relationship information between roads is used to improve the map data update efficiency and map product quality. It integrates the POI point matching algorithm, the POI point surrounding road network matching algorithm, and the POI point surrounding road network similarity comparison algorithm, and obtains the POI point surrounding road network similarity and POI point matching results at the same time, which can provide more reference information for manual operations, improve the POI point matching accuracy, and effectively reduce the workload of manual data verification and data update. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0045] Figure 1 It is a flow chart of a POI data correction method based on image segmentation in one embodiment of the present invention;

[0046] Figure 2 is a schematic diagram of a training image pair constructed for each POI point in an embodiment of the present invention;

[0047] Figure 3 Schematic diagram of a network architecture of a preset POI deviation correction matching model in one embodiment of the present invention;

[0048] Figure 4 It is a structural schematic diagram of a POI data deviation correction device based on image segmentation in one embodiment of the present invention;

[0049] Figure 5 is a schematic diagram of a computer device in one embodiment of the present invention. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] In one embodiment, if Figure 1 As shown, a POI data correction method based on image segmentation is provided, comprising the following steps:

[0052] In step S110, a training image data set is constructed, wherein the training image data set includes A-source road network image data and B-source road network image data;

[0053] In the embodiment of the present application, the A-source road network image data is derived from the A-source road network training data, and the A-source road network may refer to a self-owned road network. The B-source road network image data is derived from the B-source road network training data, and the B-source road network may refer to an open-source road network. The self-owned road network data and the open-source road network data may be specifically historical map data extracted from the corresponding road network, and may include information such as buildings, roads, routes, longitude and latitude.

[0054] It should be noted that the A-source road network image data may include the first A-source road network image training data within the M-meter range around the POI point to be matched, and the second A-source road network image training data within the N-meter range around the POI point to be matched. The B-source road network image data includes the first B-source road network image training data corresponding to the first A-source road network image training data, and the second B-source road network image training data corresponding to the second A-source road network image training data, M>N.

[0055] For example, see Figure 2 , provides a training image pair corresponding to each POI point, as well as the corresponding POI point position and surrounding road network prediction results in the open source road network obtained after image segmentation based on the training image pair. The gray origin in the figure represents the POI point position. Taking a POI point in the proprietary road network as an example, the surrounding road network data within 50 meters and 100 meters around the POI point are respectively intercepted in the proprietary road network data, and the surrounding road network data corresponding to the surrounding road network data intercepted in the open source road network data is intercepted. The above-mentioned surrounding road network data are respectively generated into image data, and the POI point is drawn into the image to realize the annotation of the POI point, and the image pair of the POI point can be obtained. For each POI point, an image pair can be constructed in the above manner, and image pairs of multiple POI points can be obtained to construct a training data set and a test data set.

[0056] Optionally, the marking method for POI points can be specifically as follows: if the surrounding road network of the open source data is relatively similar to the surrounding road network of the own road network, then the position of the surrounding road network corresponding to the POI point and the position corresponding to the POI point are marked, and marked as similar road category 1; if the surrounding road network of the open source data is significantly different from the surrounding road network of the own road network, and the corresponding POI point position cannot be determined, then the corresponding surrounding road network is marked, the POI point position does not need to be marked, and it is marked as different road category 0.

[0057] In step S120, based on the A-source road network image data and the B-source road network image data, the preset POI deviation correction matching model is iteratively trained, and when the preset convergence condition is met, a trained POI deviation correction matching model is obtained;

[0058] In the embodiment of the present application, the preset POI deviation correction matching model may specifically include a POI point matching algorithm, a POI point surrounding road network matching algorithm, and a POI point surrounding road network similarity comparison algorithm. For the input A source road network image data and B source road network image data, firstly, the input image data may be subjected to image superposition, feature extraction, feature splicing and other operations, and then the POI point and the surrounding road network are matched through the POI point matching algorithm and the POI point surrounding road network matching algorithm, such as feature decoding and pixel classification, to obtain the POI point matching result. At the same time, the surrounding road network similarity algorithm, such as cosine similarity, Euclidean distance, intersection-over-union algorithm, etc., may be used to calculate the surrounding road network similarity between the own road network and the open source road network to obtain the surrounding road network similarity comparison result. Based on the POI point matching result and the surrounding road network similarity comparison result, the matching loss value and the similarity loss value are calculated respectively, and based on the loss value, the parameters of the preset POI correction matching model are adjusted, and the above steps are repeated until the preset convergence conditions are met, such as the number of iterations reaches the preset number, or the loss value is less than the preset threshold, a trained POI correction matching model can be obtained.

[0059] It should be noted that the binary cross entropy loss function BCE Loss can be used to calculate the loss value in the POI and surrounding road network matching training part, and the contrast loss function ContrastiveLoss can be used to calculate the loss value in the similarity comparison module training part. The similarity calculation can use Euclidean distance, IOU calculation, etc. After the algorithm model is built, the model can be trained and tested based on the constructed training and test sample sets to achieve optimal model preservation.

[0060] It should be noted that the trained POI deviation correction matching model can also be tested and evaluated based on the test data set constructed above, so as to continuously optimize the model and improve the prediction accuracy.

[0061] In step S130, the data to be detected is input into the trained preset POI deviation correction matching model to obtain the POI point matching result and the surrounding road network similarity comparison result;

[0062] In an embodiment of the present application, the data to be detected may specifically include first A-source road network image data within M meters around the POI point to be matched in the A-source road network, second A-source road network image data within N meters around the POI point to be matched, and first B-source road network image data corresponding to the first A-source road network image data, where M>N. Exemplarily, the surrounding road network data within 50 meters and 100 meters around the POI point can be intercepted in the A-source road network with the POI point to be matched as the center, and the surrounding road network data corresponding to the surrounding road network data within 100 meters around the POI point intercepted in the open source road network data can be intercepted. It should be noted that the A-source road network data and the B-source road network data can also be other different road network data, and road network data from different sources can be collected according to actual reasoning needs.

[0063] In the inference stage, the above surrounding road network data are converted into image data respectively and input into the preset POI correction matching model that has been trained. The model can perform image superposition, feature extraction, feature splicing and other operations on the input image data respectively, and then match the POI point and the surrounding road network through the POI point matching algorithm and the POI point surrounding road network matching algorithm, such as feature decoding and pixel classification, to obtain the POI point matching result corresponding to the data to be detected. At the same time, the surrounding road network similarity algorithm, such as cosine similarity, Euclidean distance, intersection-over-union algorithm, etc., can be used to calculate the surrounding road network similarity between the own road network and the open source road network to obtain the surrounding road network similarity comparison result corresponding to the data to be detected.

[0064] In step S140, based on the POI point matching result and the surrounding road network similarity comparison result, the POI points in the road network to be updated are updated.

[0065] In the embodiment of the present application, based on the similarity comparison result of the surrounding road network, if the similarity value is greater than a preset similarity threshold, such as 90%, the POI point matching result is retained, and the POI point position in the road network to be updated is updated based on the matching result, otherwise, through manual authentication, the POI point position in the road network to be updated is updated based on the manual authentication result. It should be noted that the road network to be updated refers to the B source road network data in the above-mentioned data to be detected.

[0066] In one embodiment of the present application, a POI data correction method based on image segmentation is provided, including: constructing a training image data set, the training image data set including A-source road network image data and B-source road network image data; iteratively training a preset POI correction matching model based on the A-source road network image data and the B-source road network image data, and obtaining a trained POI correction matching model when the preset convergence condition is met; inputting the data to be detected into the trained preset POI correction matching model to obtain a POI point matching result and a surrounding road network similarity comparison result; based on the POI point matching result and the surrounding road network similarity comparison result, updating the POI points in the road network to be updated. In the embodiment of the present application, an image data set is constructed based on multi-source road network data, and the relative geometric position relationship information between roads is used to improve the map data update efficiency and map product quality. It integrates the POI point matching algorithm, the POI point surrounding road network matching algorithm, and the POI point surrounding road network similarity comparison algorithm, and obtains the POI point surrounding road network similarity and POI point matching results at the same time, which can provide more reference information for manual operations, improve the POI point matching accuracy, and effectively reduce the workload of manual data verification and data update.

[0067] In one embodiment of the present application, the data to be detected is input into the trained preset POI deviation correction matching model to obtain the POI point matching result and the surrounding road network similarity comparison result, including:

[0068] Superimposing the first A-source road network image data and the second A-source road network image data, and extracting features from the superimposed image data to obtain an A-source road network feature vector;

[0069] Based on the B-source road network image data and the second A-source road network image data, a sampling feature vector and a splicing feature vector are obtained;

[0070] Perform feature decoding and pixel point classification on the A-source road network feature vector and the spliced ​​feature vector to obtain the POI point matching result;

[0071] Based on the POI point matching result and the sampling feature vector, the similarity between the peripheral road network of the A source road network and the peripheral road network of the B source road network is calculated to obtain the peripheral road network similarity comparison result.

[0072] See also Figure 3, the first A-source road network image data and the second A-source road network image data in the A-source road network are superimposed by image superposition algorithms such as affine transformation and image distortion. For example, the surrounding road network images within a range of 100 meters around the POI point to be matched and the surrounding road network images within a range of 50 meters are superimposed, and the superimposed images are input into a feature extraction network, such as a convolutional neural network CNN for feature extraction to obtain a feature vector matrix of the A-source road network. For the B-source road network image data, it can be a surrounding road network image in an area corresponding to the surrounding road network image within a range of 100 meters around the POI point to be matched. It is also subjected to feature extraction and can be spliced ​​with the features obtained by sampling the second A-source road network image data, that is, the surrounding road network image within a range of 50 meters. Then, the spliced ​​feature vectors and the feature vector matrix of the A-source road network are input into the decoder together for feature decoding and pixel classification to output the POI point matching result. The decoded features can be compared with the features sampled from the second A-source road network image data to calculate the similarity to obtain the surrounding road network similarity comparison result.

[0073] Further, obtaining a sampling feature vector and a splicing feature vector based on the B-source road network image data and the second A-source road network image data includes:

[0074] Extracting image features from the B-source road network image data to obtain a B-source road network feature vector;

[0075] Up-sampling / down-sampling the second A-source road network image data to obtain the sampling feature vector;

[0076] The B-source road network feature vector is concatenated with the sampling feature vector to obtain a concatenated feature vector.

[0077] Optionally, feature extraction can be performed on the B-source road network image data through a feature extraction network, such as a convolutional neural network CNN, to obtain a B-source road network feature vector. At the same time, a sampling operation can be performed on the second A-source road network image data through upsampling or downsampling to obtain a sampled feature vector. Then, the B-source road network feature vector and the sampled feature vector are feature spliced ​​through a feature splicing algorithm, such as a scale-invariant feature transform (SIFT) algorithm or a concat function to obtain a spliced ​​feature vector.

[0078] In an embodiment of the present application, the updating of the POI points in the road network to be updated based on the POI point matching result and the surrounding road network similarity comparison result includes:

[0079] Based on the surrounding road network similarity comparison result, when the surrounding road network similarity is greater than / equal to a preset similarity threshold, the POI matching result is subjected to an inverse transformation of longitude and latitude to obtain the corresponding POI point longitude and latitude information;

[0080] Update the corresponding POI point position in the road network to be updated based on the latitude and longitude information of the POI point; or

[0081] When the similarity of the surrounding road network is less than a preset similarity threshold, manual verification is performed to update the position of the POI point in the road network to be updated based on the verification result.

[0082] Optionally, the peripheral road network similarity comparison result refers to the calculation of the similarity between the input peripheral road network data of the A source road network and the peripheral road network data of the B source road network through a similarity algorithm, such as Euclidean distance, intersection-over-union (IOU) and other algorithms. That is, after the above-mentioned feature extraction, decoder processing and sampling, the decoded features can be obtained, and the features obtained by sampling the second A source road network image data are compared based on the above-mentioned similarity algorithm to calculate the similarity between the two, as the peripheral road network similarity comparison result in different road networks. The peripheral road network similarity comparison result includes a specific similarity value. When the similarity is greater than / equal to a preset similarity threshold, the POI matching result is subjected to an inverse longitude and latitude transformation to obtain the corresponding POI point longitude and latitude information, and the corresponding POI point position in the road network to be updated is updated based on the POI point longitude and latitude information. When the peripheral road network similarity is less than the preset similarity threshold, a secondary verification can be performed manually to update the POI point position in the road network to be updated based on the verification result.

[0083] The road network to be updated may be an open source road network, such as a B-source road network.

[0084] In one embodiment of the present application, constructing a training image data set includes:

[0085] Collect A-source road network training data, and extract first A-source road network training data within M meters around each POI point and second A-source road network training data within N meters around each POI point from the A-source road network training data, with M>N;

[0086] Converting the first A-source road network training data into first A-source road network image training data, and converting the first A-source road network training data into second A-source road network image training data;

[0087] Collecting B-source road network training data, and extracting first B-source road network training data corresponding to the first A-source road network training data and second B-source road network training data corresponding to the second A-source road network training data from the open source road network training data;

[0088] Converting the first B-source road network training data into first B-source road network image training data, and converting the second B-source road network training data into second B-source road network image training data;

[0089] The first A-source road network image training data, the second A-source road network image training data, the first B-source road network image training data, and the second B-source road network image training data are used as image pairs corresponding to POI points;

[0090] The training image data set is constructed based on the image pairs of each of the POI points.

[0091] Optionally, the A-source road network may refer to a self-owned road network, and the B-source road network may refer to an open-source road network, and data collection can be performed by connecting to each road network port. The A-source road network image training data includes the first A-source road network image training data within the M-meter range around the POI point to be matched, and the second A-source road network image training data within the N-meter range around the POI point to be matched; the B-source road network image training data includes the first B-source road network image training data corresponding to the first A-source road network image training data, and the second B-source road network image training data corresponding to the second A-source road network image training data, M>N. Exemplarily, a POI point in the self-owned road network can be taken as an example, and the surrounding road network data within 50 meters and 100 meters around the POI point are respectively intercepted in the self-owned road network data, and the surrounding road network data corresponding to the surrounding road network data intercepted in the self-owned road network data are intercepted in the open-source road network data. The above-mentioned surrounding road network data are converted into image data respectively, and the POI point is drawn into the image to realize the annotation of the POI point, and the image pair of the POI point can be obtained. For each POI point, an image pair can be constructed in the above manner, and image pairs of multiple POI points can be obtained to construct a training data set.

[0092] In an embodiment of the present application, the method further includes:

[0093] Performing a similarity comparison on the A-source road network image data and the B-source road network image data to obtain a comparison similarity;

[0094] When the comparison similarity is greater than / equal to a preset threshold, the position of the POI point to be matched, the position of the road network around the POI point to be matched, and the similar road category are marked in the B-source road network image data;

[0095] When the comparison similarity is less than a preset threshold, the matchable road network positions around the POI point to be matched and different road categories are marked in the B-source road network image data.

[0096] Optionally, after obtaining the A-source road network image data and the B-source road network image data, a similarity comparison can be performed through a similarity algorithm or a manual verification method. When the similarity is greater than or equal to a preset threshold, the POI point position can be marked. The marking method for the POI point can be specifically as follows: if the surrounding road network of the open source data is similar to the surrounding road network of the own road network, the position of the surrounding road network corresponding to the POI point and the position corresponding to the POI point are marked, and marked as similar road category 1; if the surrounding road network of the open source data is significantly different from the surrounding road network of the own road network, and the corresponding POI point position cannot be determined, the corresponding surrounding road network is marked, the POI point position does not need to be marked, and it is marked as a different road category 0.

[0097] In the embodiment of the present application, an image data set is constructed based on multi-source road network data, and the relative geometric position relationship information between roads is used to improve the efficiency of map data update and the quality of map products. The POI point matching algorithm, the POI point surrounding road network matching algorithm, and the POI point surrounding road network similarity comparison algorithm are integrated to obtain the POI point surrounding road network similarity and POI point matching results at the same time, which can provide more reference information for manual operations, improve the POI point matching accuracy, and effectively reduce the workload of manual data verification and data update.

[0098] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0099] In one embodiment, a POI data correction device based on image segmentation is provided, and the POI data correction device based on image segmentation corresponds to the POI data correction method based on image segmentation in the above embodiment. Figure 4 As shown, the POI data correction device based on image segmentation includes a training image data set construction unit 10, a model training unit 20, a prediction unit 30 and an update unit 40. Each functional module is described in detail as follows:

[0100] A training image data set construction unit 10 is used to construct a training image data set, wherein the training image data set includes A-source road network image data and B-source road network image data;

[0101] The model training unit 20 iteratively trains the preset POI deviation correction matching model based on the A-source road network image data and the B-source road network image data, and obtains a trained POI deviation correction matching model when a preset convergence condition is met;

[0102] The prediction unit 30 is used to input the data to be detected into the trained preset POI deviation correction matching model to obtain the POI point matching result and the surrounding road network similarity comparison result;

[0103] The updating unit 40 is used to update the POI points in the road network to be updated based on the POI point matching result and the surrounding road network similarity comparison result.

[0104] In one embodiment of the present application, the data to be detected includes first A source road network image data within M meters around the POI point to be matched in the A source road network, second A source road network image data within N meters around the POI point to be matched, and first B source road network image data corresponding to the first A source road network image data, M>N.

[0105] In an embodiment of the present application, the prediction unit 30 is further configured to:

[0106] Superimposing the first A-source road network image data and the second A-source road network image data, and extracting features from the superimposed image data to obtain an A-source road network feature vector;

[0107] Based on the B-source road network image data and the second A-source road network image data, a sampling feature vector and a splicing feature vector are obtained;

[0108] Perform feature decoding and pixel point classification on the A-source road network feature vector and the spliced ​​feature vector to obtain the POI point matching result;

[0109] Based on the POI point matching result and the sampling feature vector, the similarity between the peripheral road network of the A source road network and the peripheral road network of the B source road network is calculated to obtain the peripheral road network similarity comparison result.

[0110] In an embodiment of the present application, the prediction unit 30 is further configured to:

[0111] Extracting image features from the B-source road network image data to obtain a B-source road network feature vector;

[0112] Up-sampling / down-sampling the second A-source road network image data to obtain a sampling feature vector;

[0113] The B-source road network feature vector is concatenated with the sampling feature vector to obtain a concatenated feature vector.

[0114] In an embodiment of the present application, the updating unit 40 is further configured to:

[0115] Based on the surrounding road network similarity comparison result, when the surrounding road network similarity is greater than / equal to a preset similarity threshold, the POI matching result is subjected to an inverse transformation of longitude and latitude to obtain the corresponding POI point longitude and latitude information;

[0116] Update the corresponding POI point position in the road network to be updated based on the latitude and longitude information of the POI point; or

[0117] When the similarity of the surrounding road network is less than a preset similarity threshold, manual verification is performed to update the position of the POI point in the road network to be updated based on the verification result.

[0118] In one embodiment of the present application, the training image data set construction unit 10 is further used to:

[0119] Collect A-source road network training data, and extract first A-source road network training data within M meters around each POI point and second A-source road network training data within N meters around each POI point from the A-source road network training data, with M>N;

[0120] Converting the first A-source road network training data into first A-source road network image training data, and converting the first A-source road network training data into second A-source road network image training data;

[0121] Collecting B-source road network training data, and extracting first B-source road network training data corresponding to the first A-source road network training data and second B-source road network training data corresponding to the second A-source road network training data from the open source road network training data;

[0122] Converting the first B-source road network training data into first B-source road network image training data, and converting the second B-source road network training data into second B-source road network image training data;

[0123] The first A-source road network image training data, the second A-source road network image training data, the first B-source road network image training data, and the second B-source road network image training data are used as image pairs corresponding to POI points;

[0124] The training image data set is constructed based on the image pairs of each of the POI points.

[0125] In an embodiment of the present application, the method further includes:

[0126] Performing a similarity comparison on the A-source road network image data and the B-source road network image data to obtain a comparison similarity;

[0127] When the comparison similarity is greater than / equal to a preset threshold, the position of the POI point to be matched, the position of the road network around the POI point to be matched, and the similar road category are marked in the B-source road network image data;

[0128] When the comparison similarity is less than a preset threshold, the matchable road network positions around the POI point to be matched and different road categories are marked in the B-source road network image data.

[0129] In the embodiment of the present application, an image data set is constructed based on multi-source road network data, and the relative geometric position relationship information between roads is used to improve the efficiency of map data update and the quality of map products. The POI point matching algorithm, the POI point surrounding road network matching algorithm, and the POI point surrounding road network similarity comparison algorithm are integrated to obtain the POI point surrounding road network similarity and POI point matching results at the same time, which can provide more reference information for manual operations, improve the POI point matching accuracy, and effectively reduce the workload of manual data verification and data update.

[0130] For the specific definition of the POI data correction device based on image segmentation, please refer to the definition of the POI data correction method based on image segmentation above, which will not be repeated here. Each module in the above-mentioned POI data correction device based on image segmentation can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0131] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 5 As shown. The computer device includes a processor, a memory, a network interface and a database connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a readable storage medium and an internal memory. The readable storage medium stores an operating system, computer-readable instructions and a database. The internal memory provides an environment for the operation of the operating system and computer-readable instructions in the readable storage medium. The database of the computer device is used to store data involved in the POI data correction method based on image segmentation. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer-readable instructions are executed by the processor, a POI data correction method based on image segmentation is implemented. The readable storage medium provided in this embodiment includes a non-volatile readable storage medium and a volatile readable storage medium.

[0132] In one embodiment, a computer device is provided, including a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor. When the processor executes the computer-readable instructions, the steps of the above-mentioned POI data correction method based on image segmentation are implemented.

[0133] In one embodiment, a computer-readable storage medium of computer-readable instructions is provided, and the readable storage medium provided in this embodiment includes a non-volatile readable storage medium and a volatile readable storage medium. The readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by one or more processors, the steps of the above-mentioned POI data correction method based on image segmentation are implemented.

[0134] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing related hardware through computer-readable instructions, and the computer-readable instructions can be stored in a non-volatile readable storage medium or a volatile readable storage medium. When the computer-readable instructions are executed, they may include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0135] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0136] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A POI data correction method based on image segmentation, characterized in that: The method comprises: Constructing a training image data set, wherein the training image data set includes A-source road network image data and B-source road network image data; Based on the A-source road network image data and the B-source road network image data, iteratively train the preset POI deviation correction matching model, and when the preset convergence condition is met, obtain a trained POI deviation correction matching model; Input the data to be detected into the trained preset POI deviation correction matching model to obtain the POI point matching result and the surrounding road network similarity comparison result; Based on the POI point matching result and the surrounding road network similarity comparison result, the POI points in the road network to be updated are updated.

2. The POI data correction method based on image segmentation according to claim 1, characterized in that: The data to be detected includes first A-source road network image data within M meters around the POI point to be matched in the A-source road network, second A-source road network image data within N meters around the POI point to be matched, and first B-source road network image data corresponding to the first A-source road network image data, M>N.

3. The POI data correction method based on image segmentation as claimed in claim 2, characterized in that: The step of inputting the data to be detected into the trained preset POI deviation correction matching model to obtain the POI point matching result and the surrounding road network similarity comparison result includes: Superimposing the first A-source road network image data and the second A-source road network image data, and extracting features from the superimposed image data to obtain an A-source road network feature vector; Based on the B-source road network image data and the second A-source road network image data, a sampling feature vector and a splicing feature vector are obtained; Perform feature decoding and pixel point classification on the A-source road network feature vector and the spliced ​​feature vector to obtain the POI point matching result; Based on the POI point matching result and the sampling feature vector, the similarity between the peripheral road network of the A source road network and the peripheral road network of the B source road network is calculated to obtain the peripheral road network similarity comparison result.

4. The POI data correction method based on image segmentation as described in claim 3 is characterized in that: The obtaining of a sampling feature vector and a splicing feature vector based on the B-source road network image data and the second A-source road network image data includes: Extracting image features from the B-source road network image data to obtain a B-source road network feature vector; Up-sampling / down-sampling the second A-source road network image data to obtain the sampling feature vector; The B-source road network feature vector is concatenated with the sampling feature vector to obtain the concatenated feature vector.

5. The POI data correction method based on image segmentation according to claim 1, characterized in that: The updating of the POI points in the road network to be updated based on the POI point matching result and the surrounding road network similarity comparison result includes: Based on the surrounding road network similarity comparison result, when the surrounding road network similarity is greater than / equal to a preset similarity threshold, the POI matching result is subjected to an inverse transformation of longitude and latitude to obtain the corresponding POI point longitude and latitude information; Update the corresponding POI point position in the road network to be updated based on the latitude and longitude information of the POI point; or When the similarity of the surrounding road network is less than a preset similarity threshold, manual verification is performed to update the position of the POI point in the road network to be updated based on the verification result.

6. The POI data correction method based on image segmentation according to claim 1, characterized in that: Build a training image dataset, including: Collect A-source road network training data, and extract first A-source road network training data within M meters around each POI point and second A-source road network training data within N meters around each POI point from the A-source road network training data, with M>N; Converting the first A-source road network training data into first A-source road network image training data, and converting the first A-source road network training data into second A-source road network image training data; Collecting B-source road network training data, and extracting first B-source road network training data corresponding to the first A-source road network training data and second B-source road network training data corresponding to the second A-source road network training data from the open source road network training data; Converting the first B-source road network training data into first B-source road network image training data, and converting the second B-source road network training data into second B-source road network image training data; The first A-source road network image training data, the second A-source road network image training data, the first B-source road network image training data, and the second B-source road network image training data are used as image pairs corresponding to POI points; The training image data set is constructed based on the image pairs of each of the POI points.

7. The POI data correction method based on image segmentation as claimed in claim 1, characterized in that: The method further comprises: Performing a similarity comparison on the A-source road network image data and the B-source road network image data to obtain a comparison similarity; When the comparison similarity is greater than / equal to a preset threshold, the position of the POI point to be matched, the position of the road network around the POI point to be matched, and the similar road category are marked in the B-source road network image data; When the comparison similarity is less than a preset threshold, the matchable road network positions around the POI point to be matched and different road categories are marked in the B-source road network image data.

8. A POI data correction device based on image segmentation, characterized in that: The device comprises: A training image data set construction unit, used to construct a training image data set, wherein the training image data set includes A-source road network image data and B-source road network image data; A model training unit, based on the A-source road network image data and the B-source road network image data, iteratively trains the preset POI deviation correction matching model, and obtains a trained POI deviation correction matching model when a preset convergence condition is met; A prediction unit, used for inputting the data to be detected into the trained preset POI deviation correction matching model to obtain the POI point matching result and the surrounding road network similarity comparison result; The updating unit is used to update the POI points in the road network to be updated based on the POI point matching result and the surrounding road network similarity comparison result.

9. A computer device comprising a memory, a processor, and computer-readable instructions stored in the memory and executed on the processor, characterized in that: When the processor executes the computer-readable instructions, the POI data correction method based on image segmentation as described in any one of claims 1 to 7 is implemented.

10. A readable storage medium having computer readable instructions stored thereon, characterized in that: When the computer-readable instructions are executed by a processor, the POI data correction method based on image segmentation as described in any one of claims 1 to 7 is implemented.

Images