Map auditing method and device based on big data platform

Abstract

The embodiment of the invention discloses a map auditing method and device based on a big data platform, and relates to the technical field of map auditing. The method comprises the following steps: acquiring map data to be audited; inputting the to-be-audited map data into an auditing element extraction network to obtain an element feature set containing a plurality of auditing elements; inputting each element feature in the element feature set into a corresponding element auditing branch network to obtain a branch result output by each element auditing branch network; and finally, based on a plurality of branch results output by the plurality of element auditing branch networks, obtaining an auditing result of the map data to be audited. Through the map auditing method and device based on the big data platform disclosed by the embodiment of the specification, the efficiency and accuracy of map auditing work can be greatly improved, and meanwhile, the labor cost spent in the auditing process is reduced.

Description

Technical Field

[0001] The present application relates to the field of map review technology, and more specifically, to a map review method and device based on a big data platform. Background Art

[0002] Maps abstract, summarize, model and express relevant natural and social phenomena in a symbolic form. They are not only a carrier of spatial information, but also an important means for people to understand the spatial distribution and pattern of the real world.

[0003] Map review refers to the examination and verification of maps that are public or to be public, thereby reducing the generation of problematic maps, ensuring the accuracy and compliance of map content, and safeguarding national security and interests.

[0004] Generally speaking, map review involves strict control of key elements of the map, such as geographic information, borders, road networks, place names, special places, etc., as well as comprehensive supervision of map production, publication, and display. However, at present, map review mainly adopts manual review method, relying on map review experts to use labor-intensive human sea tactics for visual interpretation. This method is not only inefficient and labor-intensive, but also has strong subjective factors, which may lead to inaccurate review results.

[0005] Based on this, it is necessary to study a more scientific map review method to improve the efficiency and accuracy of map review work, while reducing the manpower cost of the review process. Summary of the invention

[0006] To solve the above problems, one aspect of an embodiment of this specification provides a map review method based on a big data platform, the method comprising:

[0007] Get map data to be reviewed;

[0008] Inputting the map data to be reviewed into a review element extraction network to obtain an element feature set including multiple review elements;

[0009] Input each element feature in the element feature set into the corresponding element review branch network to obtain a branch result output by each element review branch network, wherein the element review branch network and the review element extraction network are deployed on a big data platform, and the review element extraction network is obtained by collaborative training with multiple training sample sets corresponding to multiple map review elements and element review branch networks corresponding to the multiple map review elements;

[0010] Based on the multiple branch results output by the multiple element review branch networks, the review results of the map data to be reviewed are obtained.

[0011] In some embodiments, the audit element extraction network includes a plurality of element extraction units, and the map data to be audited is input into the audit element extraction network to obtain an element feature set containing a plurality of audit elements, including:

[0012] Extracting element features respectively corresponding to the plurality of map review elements from the map data to be reviewed through the plurality of element extraction units in the review element extraction network;

[0013] The element feature set is obtained according to the element features respectively corresponding to the multiple map review elements.

[0014] In some embodiments, each of the feature extraction units is connected to a corresponding feature review branch network, and each feature feature in the feature feature set is input into the corresponding feature review branch network, including: each feature extraction unit extracts the feature from the map data to be reviewed and inputs it into the feature review branch network connected to it.

[0015] In some embodiments, before obtaining the map data to be reviewed, the method further includes:

[0016] Acquire sample data through a big data platform to obtain multiple training sample sets corresponding to the multiple map review elements, wherein the training sample set corresponding to each map review element includes positive samples and negative samples;

[0017] Determine an i-th target sample set matching the i-th target map review element from the multiple training sample sets, wherein i is a natural number greater than or equal to 1 and less than or equal to N, and N is the total number of the map review elements;

[0018] Input the j-th sample data in the i-th target sample set into the i-th element extraction unit of the audit element extraction network to obtain the sample element feature corresponding to the j-th sample data in the i-th target sample set, where j is a natural number greater than or equal to 1 and less than or equal to M, and M is the total number of sample data in the i-th target sample set;

[0019] Input the sample element feature corresponding to the j-th sample data in the i-th target sample set into the i-th element audit branch network connected to the i-th element extraction unit, and obtain the sample branch result output by the i-th element audit branch network after processing the j-th sample data in the i-th target sample set;

[0020] Determine the loss data corresponding to the i-th element extraction unit and the i-th element audit branch network based on the difference between the sample branch results corresponding to all sample data in the i-th target sample set and the reference results;

[0021] Adjusting the network parameters of the i-th element extraction unit and the i-th element audit branch network according to the loss data until the corresponding preset conditions are met, thereby completing the collaborative training of the i-th element extraction unit and the i-th element audit branch network;

[0022] When all the feature extraction units and the connected feature audit branch networks have completed training, a trained audit feature extraction network and a trained feature audit branch network are obtained.

[0023] In some embodiments, the acquiring of sample data through a big data platform to obtain a plurality of training sample sets corresponding to the plurality of map review elements respectively includes:

[0024] For each target map review element, sample map data related to the target map review element is collected from multiple data sources, wherein the sample map data includes standard map data, problematic map data uploaded by users, and problematic map data recorded in the historical review process;

[0025] Mark each sample map data according to the problem situation corresponding to the sample map data under the target map review element, and obtain a reference result corresponding to each sample map data;

[0026] The reference results are combined with corresponding sample map data to obtain a training sample set corresponding to the target map review element.

[0027] In some embodiments, the combining of the reference result with the corresponding sample map data to obtain a training sample set corresponding to the target map review element includes:

[0028] Performing sample augmentation processing based on original training samples obtained by combining the reference results with corresponding sample map data to obtain augmented training samples;

[0029] Based on the original training samples and the augmented training samples, a training sample set corresponding to the target map review element is obtained.

[0030] In some embodiments, the sample augmentation processing is performed on the original training sample obtained by combining the reference result with the corresponding sample map data to obtain the augmented training sample, including:

[0031] Acquire a first original training sample corresponding to the standard map data and a second original training sample corresponding to any problem map data;

[0032] The method further comprises: calculating a difference between the first original training sample and the second original training sample, and transforming at least part of the data in the second original training sample according to the difference between the first original training sample and the second original training sample to obtain the augmented training sample; wherein the transforming at least part of the data in the second original training sample according to the difference between the first original training sample and the second original training sample comprises amplifying the difference or reducing the difference.

[0033] In some embodiments, the determining of the loss data corresponding to the i-th element extraction unit and the i-th element audit branch network based on the difference between the sample branch results corresponding to all sample data in the i-th target sample set and the reference results includes:

[0034] The loss data corresponding to the i-th element extraction unit and the i-th element audit branch network is calculated by the following formula:

[0035]

[0036] Among them, L i represents the loss data corresponding to the i-th element extraction unit and the i-th element audit branch network; M represents the total number of sample data in the i-th target sample set; G j represents the reference result corresponding to the jth sample data in the i-th target sample set, G j ' represents the sample branch result corresponding to the jth sample data in the i-th target sample set; |G j -G j '| indicates G j With G j 'The distance between.

[0037] In some embodiments, the audit result of the map data to be audited is obtained based on the multiple branch results output by the multiple element audit branch networks, including:

[0038] Performing vector concatenation on the multiple branch results to obtain a target output vector;

[0039] The target output vector is mapped according to a preset mapping rule to obtain an audit result corresponding to the map data to be audited.

[0040] Another aspect of the embodiments of this specification further provides a map review device based on a big data platform, the device comprising:

[0041] The acquisition module is used to obtain the map data to be reviewed;

[0042] An audit element extraction module, used for inputting the map data to be audited into an audit element extraction network to obtain an element feature set including a plurality of audit elements;

[0043] An audit branch result determination module is used to input each element feature in the element feature set into a corresponding element audit branch network to obtain a branch result output by each element audit branch network, wherein the element audit branch network and the audit element extraction network are deployed on a big data platform, and the audit element extraction network is obtained by collaborative training with multiple training sample sets corresponding to multiple map audit elements and element audit branch networks corresponding to the multiple map audit elements;

[0044] The audit result determination module is used to obtain the audit result of the map data to be audited based on the multiple branch results output by the multiple element audit branch networks.

[0045] The map review method and device based on the big data platform provided in the embodiments of this specification may bring about at least the following beneficial effects:

[0046] (1) Through collaborative training, an audit element extraction network and an element audit branch network are constructed, and then multiple audit elements in the map data to be audited are extracted through the audit element extraction network. Finally, multiple element audit branch networks are used to perform multiple audits on the map data to be audited based on the multiple element features extracted by the audit element extraction network to obtain the final audit results, which can greatly improve the efficiency and accuracy of map audit work and reduce the labor cost of the audit process;

[0047] (2) By using different training sample sets to separately train the element extraction units used to extract different audit elements and the element audit branch networks used to audit different elements, the extraction and audit capabilities of each element extraction unit and element audit branch network for related elements can be improved, and the training process can be shortened;

[0048] (3) By performing augmentation processing based on the difference between the first original training sample corresponding to the standard map data and the second original training sample corresponding to the problem map data, the diversity of samples and the generalization ability of the network model can be increased without requiring a lot of manpower costs.

[0049] Additional features will be described in part in the following description. For those skilled in the art, it will become apparent by reviewing the following and accompanying drawings, or it may be understood by the production or operation of the examples. The features of this specification may be implemented and obtained by practicing or using various aspects of the methods, tools, and combinations described in the following detailed examples. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] This specification will be further described in the form of exemplary embodiments, which will be described in detail by the accompanying drawings. These embodiments are not restrictive, and in these embodiments, the same number represents the same structure, wherein:

[0051] Figure 1 is a schematic diagram of an exemplary system structure of a big data platform according to some embodiments of this specification;

[0052] Figure 2 is an exemplary module diagram of a map review device based on a big data platform according to some embodiments of this specification;

[0053] Figure 3 is an exemplary step flow chart of a map review method based on a big data platform according to some embodiments of this specification;

[0054] Figure 4 It is a schematic diagram of the execution process of the map review method based on the big data platform shown in some embodiments of this specification. DETAILED DESCRIPTION

[0055] In order to more clearly illustrate the technical solutions of the embodiments of this specification, the following is a brief introduction to the drawings required for the description of the embodiments. Obviously, the drawings described below are only some examples or embodiments of this specification. For ordinary technicians in this field, this specification can also be applied to other similar scenarios based on these drawings without creative work. Unless it is obvious from the language environment or otherwise explained, the same reference numerals in the figures represent the same structure or operation.

[0056] It should be understood that the "system", "device", "unit" and / or "module" used in this specification is a method for distinguishing different components, elements, parts, parts or assemblies at different levels. However, if other words can achieve the same purpose, the words can be replaced by other expressions.

[0057] As shown in this specification and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "comprise" and "include" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0058] Flowcharts are used in this specification to illustrate the operations performed by the system according to the embodiments of this specification. It should be understood that the preceding or following operations are not necessarily performed precisely in order. Instead, the steps may be processed in reverse order or simultaneously. At the same time, other operations may also be added to these processes, or one or more operations may be removed from these processes.

[0059] The map review method and device based on the big data platform provided in the embodiments of this specification are described in detail below in conjunction with the accompanying drawings.

[0060] The big data platform is a comprehensive platform that provides functions such as big data storage, processing, analysis and visualization. For map review, by using the big data platform, it can not only reduce the workload of manual review, but also greatly improve the quality of review efficiency and ensure the accuracy and compliance of map information.

[0061] Figure 1 is a schematic diagram of an exemplary system structure of a big data platform according to some embodiments of this specification. Figure 1 In the embodiment of the present application, the big data platform 100 may include a storage device 110, a processing device 120, and an input / output interface 130. The various components in the big data platform 100 may be connected in a variety of ways. For example, they may be connected through a network and / or a bus.

[0062] Among them, the storage device 110 can be used to store data, instructions and / or any other information. In some embodiments, the storage device 110 can store data uploaded by the user through the input / output interface 130 and data processed by the processing device 120. For example, the storage device 110 can store map data uploaded by the user; for another example, the storage device 110 can store the audit results obtained by the processing device 120 for auditing the map data to be audited and the intermediate data generated during the audit process. In some embodiments, the storage device 110 may include a map audit device (i.e., a map audit device 200 based on a big data platform referred to later), which can be understood as data and / or instructions used by the processing device 120 to execute or use to complete the exemplary method described in this specification.

[0063] In the embodiment of the present application, the storage device 110 may include a large capacity memory, a removable memory, a volatile read-write memory, a read-only memory (ROM), etc. or any combination thereof. Exemplary large capacity memories may include magnetic disks, optical disks, solid-state disks, etc. In some embodiments, the storage device 110 may include a cloud storage device.

[0064] The processing device 120 can be used to perform data processing tasks, such as the collection, analysis, processing and storage of map data. The processing device 120 may include one or more processors, which may be a central processing unit (CPU), a graphics processing unit (GPU), a digital signal processor (DSP), or other types of processors. In an embodiment of the present application, the processing device 120 may be a single server or a server group, wherein the server group may be centralized or distributed. In some embodiments, the processing device 120 may obtain pre-stored computer instructions from the storage device 110 and execute the computer instructions to implement the map review method based on the big data platform described in this specification.

[0065] The input / output interface 130 may provide an interface for users to interact with the big data platform 100. Users may upload data, download data, query data, and receive processing results through the input / output interface 130. The input / output interface 130 may include a graphical user interface (GUI), a command line interface (CLI), an application program interface (API), etc. In some embodiments, the input / output interface 130 may also provide an interface with other systems to facilitate data exchange and integration between the big data platform 100 and other systems.

[0066] It should be noted that the above description of the big data platform 100 is only for example and explanation, and does not limit the scope of application of this specification. For those skilled in the art, various modifications and changes can be made to the big data platform 100 under the guidance of this specification. However, these modifications and changes are still within the scope of this specification. For example, the big data platform 100 may include more or fewer functional components.

[0067] Figure 2 is a schematic diagram of a module of a map review device based on a big data platform according to some embodiments of this specification. In some embodiments, Figure 2 The map review device 200 based on the big data platform shown can be applied to Figure 1 The big data platform 100 shown, for example, can be configured to the storage device 110 in the form of software, or configured to the processing device 120 in the form of software + hardware, so as to be used for reviewing the map data to be reviewed.

[0068] Reference Figure 2In an embodiment of the present application, a map review device 200 based on a big data platform may include an acquisition module 210, a review element extraction module 220, a review branch result determination module 230, and a review result determination module 240.

[0069] The acquisition module 210 may be used to acquire map data to be reviewed.

[0070] The audit element extraction module 220 can be used to input the map data to be audited into the audit element extraction network to obtain an element feature set including multiple audit elements.

[0071] The audit branch result determination module 230 can be used to input each element feature in the element feature set into the corresponding element audit branch network, and obtain the branch result output by each element audit branch network.

[0072] The audit result determination module 240 may be used to obtain the audit result of the map data to be audited based on the multiple branch results output by the multiple element audit branch networks.

[0073] For more details about the above modules, please refer to other places in this manual (for example Figure 3-4 part and its related description), which will not be repeated here.

[0074] It should be understood that Figure 2 The map audit device 200 based on the big data platform and its modules shown can be implemented in various ways. For example, in some embodiments, the system and its modules can be implemented by hardware, software, or a combination of software and hardware. Among them, the hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or a dedicated design hardware. It can be understood by those skilled in the art that the above methods and devices can be implemented using computer executable instructions and / or included in a processor control code, for example, such as a carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. Such code is provided on the system and its modules of this specification. Not only can the hardware circuits such as ultra-large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc. be implemented, they can also be implemented by software executed by various types of processors, and can also be implemented by a combination of the above hardware circuits and software (for example, firmware).

[0075] It should be noted that the above description of the map review device 200 based on the big data platform is provided for illustrative purposes only and is not intended to limit the scope of this specification. It is understandable that for those skilled in the art, according to the description of this specification, without departing from this principle, the various modules can be combined arbitrarily, or a subsystem can be formed to connect with other modules. For example, Figure 2 The acquisition module 210, the audit element extraction module 220, the audit branch result determination module 230 and the audit result determination module 240 described in the system may be different modules in one system, or one module may realize the functions of two or more of the above modules. For another example, the map audit device 200 based on the big data platform may also include a preprocessing module, which may be used to preprocess the aforementioned map data to be processed (e.g., denoising, error correction, enhancement, standardization, etc.). Such variations are all within the scope of protection of this specification.

[0076] Figure 3 is an exemplary step flow chart of a map review method based on a big data platform according to some embodiments of this specification. In the embodiments of the present application, the method can be performed by processing logic, which can include hardware (e.g., circuits, dedicated logic, programmable logic, microcode, etc.), software (instructions running on a processing device to perform hardware simulation), etc. or any combination thereof. In some embodiments, Figure 3 One or more operations in the flowchart of the map review method based on the big data platform can be performed by Figure 1 The method is implemented by the processing device 120 shown. For example, the method may be stored in the storage device 110 in the form of a computer program and / or instruction, and called and / or executed by the processing device 120.

[0077] Reference Figure 3 In an embodiment of the present application, the map review method based on the big data platform may include:

[0078] Step S310 , obtaining map data to be reviewed. In some embodiments, step S310 may be performed by the obtaining module 210 .

[0079] In the embodiment of the present application, the user can upload the map data to be reviewed to the big data platform 100 through the above-mentioned input / output interface 130, and store it in the storage device 110. Further, the acquisition module 210 can obtain the map data to be reviewed from the storage device 110, and send it to the processing device 120 for processing, so as to review its accuracy and compliance.

[0080] In the embodiment of the present application, the map data to be reviewed may be a map file in vector format or an image file in raster format. These map data may include geographic information such as roads, buildings, water bodies, vegetation, etc. In some cases, the map data may also contain specific geographic marks or annotations, which are used to indicate the attributes and / or characteristics of a specific place or area (such as place names, landforms, road names, place names, etc.).

[0081] Step S320 , inputting the map data to be audited into the audit element extraction network to obtain an element feature set including multiple audit elements. In some embodiments, step S320 may be performed by the audit element extraction module 220 .

[0082] In an embodiment of the present application, the map data to be reviewed can be input into a review element extraction network to extract the review elements in the map data to be reviewed. The review elements can be understood as information that needs to be paid attention to during the map review process. As an example only, in some embodiments of the present application, the review elements can include but are not limited to geographic information, boundary lines, road networks, place names, special places (such as confidential places), etc.

[0083] Figure 4 Schematic diagram of the execution process of the map review method based on the big data platform according to some embodiments of this specification. Figure 4 In the embodiment of the present application, the audit factor extraction network may include multiple factor extraction units (eg Figure 4 The first element extraction unit to the i-th element extraction unit in the above-mentioned figure), wherein each element extraction unit can perform feature extraction on a type of audit element, thereby obtaining an element feature set including multiple audit elements based on the multiple element extraction units, and the element feature set includes element features corresponding to the multiple audit elements respectively.

[0084] In an embodiment of the present application, the multiple feature extraction units may have the same or similar structure. For example, each feature extraction unit may include an input layer, one or more hidden layers, and an output layer, wherein the input layer is used to receive map data to be reviewed, the hidden layer is used to process data and extract features, and the output layer is used to output the extracted feature features. In some embodiments, the hidden layer may include a convolutional layer and a pooling layer, wherein the convolutional layer is used to extract features of the input data, and the pooling layer is used to reduce the feature dimension to reduce the amount of calculation. In some embodiments, the aforementioned hidden layer may adopt a deep learning model such as a convolutional neural network (CNN) to improve the accuracy and efficiency of feature extraction.

[0085] In the process of performing feature feature extraction, each feature extraction unit can be trained according to its specific audit feature type. For example, the feature extraction unit for geographic information will focus on identifying and extracting features related to geographic information from map data, while the feature extraction unit for road networks will focus on extracting features of road network structures. Similarly, the feature extraction unit for map boundary lines will focus on identifying and extracting features related to boundary lines (such as national boundaries, provincial boundaries, etc.) from map data, the feature extraction unit for place names will focus on identifying and extracting features related to place names from map data, and the feature extraction unit for special places will focus on identifying and extracting features related to special places from map data. In this way, it can be ensured that each feature extraction unit can efficiently and accurately extract the features of its corresponding type of audit feature.

[0086] In the embodiment of the present application, each element feature in the element feature set may be embodied in the form of a multi-dimensional vector, and the element features obtained by different element extraction units may have the same or different dimensions.

[0087] In some embodiments of the present application, the map data to be reviewed can be segmented according to certain rules, and feature extraction can be performed for each segmented area. Then, the feature extraction obtained from each area can be integrated according to the corresponding rules, thereby obtaining the feature extraction of the map data to be reviewed by each feature extraction unit.

[0088] Step S330, input each element feature in the element feature set into the corresponding element audit branch network, and obtain the branch results output by each element audit branch network. In some embodiments, step S330 can be performed by the audit branch result determination module 230.

[0089] Continue to refer to Figure 4 In the embodiment of the present application, each element extraction unit is connected to a corresponding element audit branch network. The audit branch result determination module 230 can extract the element features from the map data to be audited by each element extraction unit and input them into the element audit branch network connected to it for analysis, thereby obtaining the branch results (such as Figure 4 The 1st branch result ~ the i-th branch result in ).

[0090] In an embodiment of the present application, the multiple factor review branch networks may have the same or different structures. For example, in some embodiments, the multiple factor review branch networks may have the same or different numbers of classifiers. It should be noted that in an embodiment of the present application, for each factor review branch network, each classifier contained therein can process the feature of the elements corresponding to an area obtained by segmentation according to the above rules, so as to obtain the output corresponding to the area. Finally, by integrating the outputs of all classifiers according to the corresponding rules, the branch results corresponding to the factor review branch network can be obtained.

[0091] In the embodiment of the present application, the element audit branch network and the audit element extraction network are both deployed on the big data platform 100. The audit element extraction network can be obtained by collaborative training with multiple training sample sets corresponding to multiple map audit elements and element audit branch networks corresponding to the multiple map audit elements. The following is a brief description of the training process of the audit element extraction network and the element audit branch network involved in the embodiment of the present application.

[0092] In an embodiment of the present application, the collaborative training process of the audit element extraction network and the element audit branch network may include the following steps:

[0093] S1. Obtain sample data through a big data platform to obtain multiple training sample sets corresponding to the multiple map review elements, wherein the training sample set corresponding to each map review element includes positive samples and negative samples.

[0094] Specifically, in an embodiment of the present application, for each target map review element, sample map data related to the target map review element can be collected from multiple data sources, and the sample map data includes standard map data, problem map data uploaded by users, and problem map data recorded in the historical review process. Among them, the standard map data can be used as a positive sample, and the problem map data uploaded by users and the problem map data recorded in the historical review process can be used as negative samples. Furthermore, each of the sample map data can be annotated according to the actual problem situation corresponding to the sample map data under the target map review element, and a reference result corresponding to each sample map data can be obtained. Finally, by combining the reference result with the corresponding sample map data, the training sample set corresponding to the target map review element can be obtained.

[0095] For example, for the audit element of geographic information, one or more standard map data (the multiple standard map data may be multiple different versions of compliant map data) may be obtained as positive samples, and then several problematic map data that have problems with the audit element of geographic information (for example, errors in the size, shape, location, etc. of the area corresponding to water bodies or mountains) but have no errors (or compliance issues) in other aspects may be obtained as negative samples.

[0096] Furthermore, each sample map data (including positive samples and negative samples) can be annotated according to the actual problem situation corresponding to each sample map data to obtain a reference result corresponding to each sample map data. Exemplarily, in some embodiments, the reference result corresponding to the positive sample can be expressed as (1, 1..., 1, 1), indicating that there are no problems in all its map areas, and the reference result corresponding to the negative sample can be expressed as (1, 1...0, 0, 1), indicating that there are problems in the area corresponding to "0". Finally, by combining each reference result with the corresponding sample map data, a training sample set corresponding to the audit element of geographic information can be obtained.

[0097] Similarly, for the audit element of the boundary line, one or more standard map data can be obtained as positive samples, and then several problem map data with problems in the audit element of the boundary line (for example, the position of the boundary line is wrong), but no errors (or compliance issues) in other aspects can be obtained as negative samples. Furthermore, each sample map data can be annotated with the actual problem situation corresponding to each sample map data to obtain the reference result corresponding to each sample map data. Finally, by combining each reference result with the corresponding sample map data, the training sample set corresponding to the audit element of the boundary line can be obtained.

[0098] For the audit element of road network, one or more standard map data can be obtained as positive samples, and then several problem map data with problems in the audit element of road network (for example, incomplete road network or wrong road network relationship) but no errors (or compliance issues) in other aspects can be obtained as negative samples. Furthermore, each sample map data can be annotated with the actual problem situation corresponding to each sample map data to obtain the reference result corresponding to each sample map data. Finally, by combining each reference result with the corresponding sample map data, the training sample set corresponding to the audit element of road network can be obtained.

[0099] For the audit element of place names, one or more standard map data can be obtained as positive samples, and then several problem map data with problems in the audit element of place names (such as wrong place names or wrong labeling positions), but no errors (or compliance issues) in other aspects can be obtained as negative samples. Furthermore, each sample map data can be labeled with the actual problem situation corresponding to each sample map data to obtain the reference result corresponding to each sample map data. Finally, by combining each reference result with the corresponding sample map data, the training sample set corresponding to the audit element of place names can be obtained.

[0100] For the audit element of special places, one or more standard map data can be obtained as positive samples, and then several problem map data with problems in the audit element of special places (such as disclosing the name and location of confidential units in the map, etc.), but no errors (or compliance issues) in other aspects can be obtained as negative samples. Furthermore, each sample map data can be annotated with the actual problem situation corresponding to each sample map data to obtain the reference result corresponding to each sample map data. Finally, by combining each reference result with the corresponding sample map data, the training sample set corresponding to the audit element of special places can be obtained.

[0101] In some embodiments of the present application, in order to increase the number of training samples and improve the generalization ability of the network model, in the process of combining the reference results with the corresponding sample map data to obtain the training sample set corresponding to the target map review element, sample augmentation processing can be performed based on the original training samples obtained by combining the reference results with the corresponding sample map data to obtain augmented training samples; then, based on the original training samples and the augmented training samples, a training sample set corresponding to the target map review element is obtained.

[0102] For example, in some embodiments, a first original training sample corresponding to the standard map data and a second original training sample corresponding to any problem map data may be obtained; then, the difference between the first original training sample and the second original training sample is calculated, and at least part of the data in the second original training sample is transformed according to the difference between the first original training sample and the second original training sample to obtain the augmented training sample.

[0103] Specifically, in the embodiment of the present application, the difference between the first original training sample and the second original training sample may refer to the difference between the sample map data. In some embodiments of the present application, the sample map data may be converted into a corresponding multi-dimensional feature vector (for example, processed by text vectorization or image vectorization, and the conversion process is a prior art), and then the difference between the first original training sample and the second original training sample may be obtained by calculating the distance between the multi-dimensional feature vectors (for example, the Euclidean distance).

[0104] In some embodiments of the present application, at least part of the data in the second original training sample can be transformed by enlarging or reducing the difference between the first original training sample and the second original training sample, thereby obtaining an augmented training sample. As an example only, for the training sample set corresponding to the audit element of place name, the wrong place name in the second original training sample can be modified in a direction close to the correct place name to reduce the difference (for example, the correct place name in the first original training sample is "Bozhou", and the wrong place name in the second original training sample is "Haozhou", then the wrong place name in the second original training sample can be changed to "Bozhou" to reduce the difference); or modified in a direction away from the correct place name to enlarge the difference (for example, the correct place name in the first original training sample is "Bozhou", and the wrong place name in the second original training sample is "Bozhou", then the wrong place name in the second original training sample can be changed to "Haozhou" to enlarge the difference).

[0105] Similarly, for training sample sets corresponding to other audit elements such as geographic information, boundary lines, road networks, special places, etc., at least part of the data of the second original training sample in the corresponding training sample set can be changed by fine-tuning, adding or reducing content, etc., so as to obtain augmented training samples of the corresponding training sample set.

[0106] Furthermore, after obtaining the training sample set corresponding to each map review element, the corresponding element extraction unit and element review branch network can be trained based on the training sample set corresponding to each map review element. Please refer to steps S2 to S6 for the specific process:

[0107] S2, determining the i-th target sample set matching the i-th target map review element from the multiple training sample sets, wherein i is a natural number greater than or equal to 1 and less than or equal to N, and N is the total number of the map review elements.

[0108] S3, inputting the j-th sample data in the i-th target sample set into the i-th element extraction unit of the audit element extraction network to obtain the sample element feature corresponding to the j-th sample data in the i-th target sample set, wherein j is a natural number greater than or equal to 1 and less than or equal to M, and M is the total number of sample data in the i-th target sample set.

[0109] S4, inputting the sample element features corresponding to the j-th sample data in the i-th target sample set into the i-th element review branch network connected to the i-th element extraction unit, and obtaining the sample branch result output by the i-th element review branch network after processing the j-th sample data in the i-th target sample set.

[0110] S5, determining the loss data corresponding to the i-th element extraction unit and the i-th element audit branch network based on the difference between the sample branch results corresponding to all sample data in the i-th target sample set and the reference results.

[0111] S6. Adjust the network parameters of the i-th element extraction unit and the i-th element review branch network according to the loss data until the corresponding preset conditions are met, thereby completing the collaborative training of the i-th element extraction unit and the i-th element review branch network.

[0112] In the embodiment of the present application, the loss data corresponding to the i-th element extraction unit and the i-th element audit branch network can be expressed by the following formula:

[0113]

[0114] Among them, L i represents the loss data corresponding to the i-th element extraction unit and the i-th element audit branch network; M represents the total number of sample data in the i-th target sample set; G j represents the reference result corresponding to the jth sample data in the i-th target sample set, G' j represents the sample branch result corresponding to the j-th sample data in the i-th target sample set (the reference result and the sample branch result can be represented in the form of a multidimensional vector); |G j -G′ j | indicates G j With G' j The distance between them (e.g. Euclidean distance).

[0115] Specifically, in an embodiment of the present application, the network parameters of the i-th element extraction unit and the i-th element audit branch network can be adjusted according to the loss data using optimization algorithms such as back propagation and gradient descent until the corresponding preset conditions (the preset conditions can be set according to actual needs, and this specification does not specifically limit them) are met, and the collaborative training of the i-th element extraction unit and the i-th element audit branch network is completed. More details about the network parameter adjustment and collaborative training can be regarded as prior art and will not be discussed in detail in this specification.

[0116] In an embodiment of the present application, through the above steps S2 to S6, the factor extraction units and factor review branch networks corresponding to M audit factors can be iteratively trained. When all factor extraction units and the connected factor review branch networks have completed training, a trained audit factor extraction network and a trained factor review branch network can be obtained.

[0117] It should be noted that in some embodiments of the present application, when the audit requirements change, the above-mentioned standard map data can be replaced with the latest standard map data, and then the audit element extraction network and the element audit branch network can be retrained, so that the audit element extraction network and the element audit branch network can obtain the ability to perform map audits according to the latest audit requirements.

[0118] Step S340 , based on the multiple branch results output by the multiple element review branch networks, obtain the review result of the map data to be reviewed. In some embodiments, step S340 may be performed by the review result determination module 240 .

[0119] In the embodiment of the present application, each branch result output by the element audit branch network can represent an audit result of one aspect. Figure 4 In the embodiment of the present application, the branch result splicing unit can perform vector splicing on the multiple branch results to obtain a target output vector; then, the mapping unit maps the target output vector according to a preset mapping rule to obtain the audit result corresponding to the map data to be audited. In the embodiment of the present application, the branch result splicing unit and the mapping unit can be implemented in software.

[0120] Specifically, the branch result output by each element audit branch network can be a multi-dimensional vector, and different dimensions can represent the problem status of different regions or different locations in the corresponding audit elements (for example, 1 can indicate that there is no problem in the corresponding region or location, and 0 indicates that there is a problem in the corresponding audit element in the corresponding region or location). In the embodiment of the present application, the target output vector is obtained by vector splicing of the multiple branch results, so that all problems of the map data to be audited can be reflected through the target output vector.

[0121] Specifically, in some embodiments of the present application, a mapping rule can be established based on the correspondence between each data in the target output vector and the map position or area, and then the target output vector is mapped using the mapping rule, so that the problem status of each area or position in the map data to be reviewed is determined based on the position of each data in the target output vector, and the final review result is obtained.

[0122] It should be pointed out that in the embodiment of the present application, multiple audit elements in the map data to be audited are extracted through the above-mentioned audit element extraction network, and then the map data to be audited is audited in multiple aspects using multiple element features extracted by the audit element extraction network through multiple element audit branch networks to obtain the final audit result, which can greatly improve the efficiency and accuracy of the map audit work and reduce the manpower cost spent in the audit process.

[0123] In summary, the beneficial effects that may be brought about by the embodiments of this specification include but are not limited to: (1) In the map review method and device based on the big data platform provided in some embodiments of this specification, a review element extraction network and a feature review branch network are constructed through collaborative training, and then multiple review elements in the map data to be reviewed are extracted through the review element extraction network. Finally, multiple feature features extracted by the review element extraction network are used to conduct multiple reviews of the map data to be reviewed through multiple feature review branch networks to obtain the final review results, which can greatly improve the efficiency and accuracy of map review work and reduce the labor cost of the review process; (2) In the map review method and device based on the big data platform provided in some embodiments of this specification, a review element extraction network and a feature review branch network are constructed through collaborative training, and then multiple review elements in the map data to be reviewed are extracted through multiple feature review branch networks. In the map review method and device based on the big data platform, by using different training sample sets to separately train the element extraction units used to extract different review elements and the element review branch networks used for reviewing different elements, the extraction and review capabilities of each element extraction unit and the element review branch network for related elements can be improved, and the training process can be shortened; (3) In the map review method and device based on the big data platform provided in some embodiments of the present specification, by performing augmentation processing based on the difference between the first original training sample corresponding to the standard map data and the second original training sample corresponding to the problem map data, the diversity of samples and the generalization ability of the network model can be increased without requiring a lot of manpower costs.

[0124] It should be noted that different embodiments may produce different beneficial effects. In different embodiments, the beneficial effects that may be produced may be any one or a combination of the above, or may be any other beneficial effects that may be obtained.

[0125] The basic concepts have been described above. Obviously, for those skilled in the art, the above detailed disclosure is only for example and does not constitute a limitation of this specification. Although not explicitly stated here, those skilled in the art may make various modifications, improvements and corrections to this specification. Such modifications, improvements and corrections are suggested in this specification, so such modifications, improvements and corrections still belong to the spirit and scope of the exemplary embodiments of this specification.

[0126] At the same time, this specification uses specific words to describe the embodiments of this specification. For example, "one embodiment", "an embodiment", and / or "some embodiments" refer to a certain feature, structure or characteristic related to at least one embodiment of this specification. Therefore, it should be emphasized and noted that "one embodiment" or "an embodiment" or "an alternative embodiment" mentioned twice or more in different positions in this specification does not necessarily refer to the same embodiment. In addition, certain features, structures or characteristics in one or more embodiments of this specification can be appropriately combined.

[0127] In addition, it will be understood by those skilled in the art that various aspects of this specification may be illustrated and described by a number of patentable categories or situations, including any new and useful process, machine, product or combination of substances, or any new and useful improvements thereto. Accordingly, various aspects of this specification may be performed entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. The above hardware or software may be referred to as "data blocks", "modules", "engines", "units", "components" or "systems". In addition, various aspects of this specification may be represented as a computer product located in one or more computer-readable media, which includes computer-readable program code.

[0128] A computer storage medium may include a propagated data signal containing computer program code, for example, in baseband or as part of a carrier wave. The propagated signal may be in a variety of forms, including electromagnetic, optical, etc., or a suitable combination. A computer storage medium may be any computer-readable medium other than a computer-readable storage medium, which can be connected to an instruction execution system, device or apparatus to communicate, propagate or transmit the program for use. The program code on the computer storage medium may be transmitted via any suitable medium, including radio, cable, fiber optic cable, RF, or similar media, or any combination of the above media.

[0129] The computer program codes required for the operation of each part of this specification can be written in any one or more programming languages, including object-oriented programming languages ​​such as Java, Scala, Smalltalj, Eiffel, JADE, Emerald, C++, C#, VB.NET, Python, etc., conventional procedural programming languages ​​such as C language, Visual Basic, Fortran2003, Perl, COBOL2002, PHP, ABAP, dynamic programming languages ​​such as Python, Ruby and Groovy, or other programming languages, etc. The program code can be run entirely on the user's computer, or run on the user's computer as an independent software package, or run partially on the user's computer and partially on a remote computer, or run entirely on a remote computer or processing device. In the latter case, the remote computer can be connected to the user's computer through any network form, such as a local area network (LAN) or a wide area network (WAN), or connected to an external computer (e.g., via the Internet), or in a cloud computing environment, or used as a service such as software as a service (SaaS).

[0130] In addition, unless explicitly stated in the claims, the order of the processing elements and sequences described in this specification, the use of alphanumeric characters, or the use of other names are not intended to limit the order of the processes and methods of this specification. Although the above disclosure discusses some invention embodiments that are currently considered useful through various examples, it should be understood that such details are only for illustrative purposes, and the attached claims are not limited to the disclosed embodiments. On the contrary, the claims are intended to cover all modifications and equivalent combinations that are consistent with the essence and scope of the embodiments of this specification. For example, although the system components described above can be implemented by hardware devices, they can also be implemented only by software solutions, such as installing the described system on an existing processing device or mobile device.

[0131] Similarly, it should be noted that in order to simplify the description disclosed in this specification and thus help understand one or more embodiments of the invention, in the above description of the embodiments of this specification, multiple features are sometimes combined into one embodiment, figure or description thereof. However, this disclosure method does not mean that the features required by the subject matter of this specification are more than the features mentioned in the claims. In fact, the features of the embodiments are less than all the features of the single embodiment disclosed above.

[0132] In some embodiments, numbers describing the number of components and attributes are used. It should be understood that such numbers used in the description of the embodiments are modified by the modifiers "about", "approximately" or "substantially" in some examples. Unless otherwise specified, "about", "approximately" or "substantially" indicate that the numbers are allowed to vary by ±20%. Accordingly, in some embodiments, the numerical parameters used in the specification and claims are approximate values, which may change according to the required features of individual embodiments. In some embodiments, the numerical parameters should take into account the specified significant digits and adopt the general method of retaining digits. Although the numerical domains and parameters used to confirm the breadth of their range in some embodiments of this specification are approximate values, in specific embodiments, the setting of such numerical values ​​is as accurate as possible within the feasible range.

[0133] Each patent, patent application, patent application publication, and other materials, such as articles, books, specifications, publications, documents, etc., cited in this specification are hereby incorporated by reference in their entirety. Except for application history documents that are inconsistent with or conflicting with the contents of this specification, documents that limit the broadest scope of the claims of this specification (currently or later attached to this specification) are also excluded. It should be noted that if the descriptions, definitions, and / or use of terms in the materials attached to this specification are inconsistent or conflicting with the contents described in this specification, the descriptions, definitions, and / or use of terms in this specification shall prevail.

[0134] Finally, it should be understood that the embodiments described in this specification are only used to illustrate the principles of the embodiments of this specification. Other variations may also fall within the scope of this specification. Therefore, as an example and not a limitation, alternative configurations of the embodiments of this specification may be considered consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to the embodiments explicitly introduced and described in this specification.

Claims

1. A map review method based on a big data platform, characterized in that: include: Get map data to be reviewed; Inputting the map data to be reviewed into a review element extraction network to obtain an element feature set including multiple review elements; Input each element feature in the element feature set into the corresponding element review branch network to obtain a branch result output by each element review branch network, wherein the element review branch network and the review element extraction network are deployed on a big data platform, and the review element extraction network is obtained by collaborative training with multiple training sample sets corresponding to multiple map review elements and element review branch networks corresponding to the multiple map review elements; Based on the multiple branch results output by the multiple element review branch networks, the review results of the map data to be reviewed are obtained.

2. The method according to claim 1, characterized in that The audit element extraction network includes a plurality of element extraction units. The map data to be audited is input into the audit element extraction network to obtain an element feature set containing a plurality of audit elements, including: Extracting element features respectively corresponding to the plurality of map review elements from the map data to be reviewed through the plurality of element extraction units in the review element extraction network; The element feature set is obtained according to the element features respectively corresponding to the multiple map review elements.

3. The method according to claim 2, characterized in that Each of the feature extraction units is connected to a corresponding feature review branch network, and each feature in the feature feature set is input into the corresponding feature review branch network, including: each feature extraction unit extracts the feature from the map data to be reviewed and inputs it into the feature review branch network connected to it.

4. The method according to claim 3, characterized in that Before obtaining the map data to be reviewed, the method further includes: Acquire sample data through a big data platform to obtain multiple training sample sets corresponding to the multiple map review elements, wherein the training sample set corresponding to each map review element includes positive samples and negative samples; Determine an i-th target sample set matching the i-th target map review element from the multiple training sample sets, wherein i is a natural number greater than or equal to 1 and less than or equal to N, and N is the total number of the map review elements; Input the j-th sample data in the i-th target sample set into the i-th element extraction unit of the audit element extraction network to obtain the sample element feature corresponding to the j-th sample data in the i-th target sample set, where j is a natural number greater than or equal to 1 and less than or equal to M, and M is the total number of sample data in the i-th target sample set; Input the sample element feature corresponding to the j-th sample data in the i-th target sample set into the i-th element audit branch network connected to the i-th element extraction unit, and obtain the sample branch result output by the i-th element audit branch network after processing the j-th sample data in the i-th target sample set; Determine the loss data corresponding to the i-th element extraction unit and the i-th element audit branch network based on the difference between the sample branch results corresponding to all sample data in the i-th target sample set and the reference results; Adjusting the network parameters of the i-th element extraction unit and the i-th element audit branch network according to the loss data until the corresponding preset conditions are met, thereby completing the collaborative training of the i-th element extraction unit and the i-th element audit branch network; When all the feature extraction units and the connected feature audit branch networks have completed training, a trained audit feature extraction network and a trained feature audit branch network are obtained.

5. The method according to claim 4, characterized in that The method of acquiring sample data through the big data platform to obtain a plurality of training sample sets corresponding to the plurality of map review elements respectively includes: For each target map review element, sample map data related to the target map review element is collected from multiple data sources, wherein the sample map data includes standard map data, problematic map data uploaded by users, and problematic map data recorded in the historical review process; Mark each sample map data according to the problem situation corresponding to the sample map data under the target map review element, and obtain a reference result corresponding to each sample map data; The reference results are combined with corresponding sample map data to obtain a training sample set corresponding to the target map review element.

6. The method according to claim 5, characterized in that The step of combining the reference result with the corresponding sample map data to obtain a training sample set corresponding to the target map review element includes: Performing sample augmentation processing based on original training samples obtained by combining the reference results with corresponding sample map data to obtain augmented training samples; Based on the original training samples and the augmented training samples, a training sample set corresponding to the target map review element is obtained.

7. The method according to claim 6, characterized in that The method of performing sample augmentation processing on the original training sample obtained by combining the reference result with the corresponding sample map data to obtain the augmented training sample includes: Acquire a first original training sample corresponding to the standard map data and a second original training sample corresponding to any problem map data; The method further comprises: calculating a difference between the first original training sample and the second original training sample, and transforming at least part of the data in the second original training sample according to the difference between the first original training sample and the second original training sample to obtain the augmented training sample; wherein the transforming at least part of the data in the second original training sample according to the difference between the first original training sample and the second original training sample comprises amplifying the difference or reducing the difference.

8. The method according to claim 4, characterized in that The determining of the loss data corresponding to the i-th element extraction unit and the i-th element audit branch network based on the difference between the sample branch results corresponding to all sample data in the i-th target sample set and the reference results includes: The loss data corresponding to the i-th element extraction unit and the i-th element audit branch network is calculated by the following formula: Among them, L i represents the loss data corresponding to the i-th element extraction unit and the i-th element audit branch network; M represents the total number of sample data in the i-th target sample set; G j represents the reference result corresponding to the jth sample data in the i-th target sample set, G' j represents the sample branch result corresponding to the jth sample data in the i-th target sample set; |G j -G' j | indicates G j With G' j The distance between.

9. The method according to claim 1, characterized in that The step of obtaining the audit result of the map data to be audited based on the multiple branch results output by the multiple element audit branch networks includes: Performing vector concatenation on the multiple branch results to obtain a target output vector; The target output vector is mapped according to a preset mapping rule to obtain an audit result corresponding to the map data to be audited.

10. A map review device based on a big data platform, characterized in that: include: The acquisition module is used to obtain the map data to be reviewed; An audit element extraction module, used for inputting the map data to be audited into an audit element extraction network to obtain an element feature set including a plurality of audit elements; An audit branch result determination module is used to input each element feature in the element feature set into a corresponding element audit branch network to obtain a branch result output by each element audit branch network, wherein the element audit branch network and the audit element extraction network are deployed on a big data platform, and the audit element extraction network is obtained by collaborative training with multiple training sample sets corresponding to multiple map audit elements and element audit branch networks corresponding to the multiple map audit elements; The audit result determination module is used to obtain the audit result of the map data to be audited based on the multiple branch results output by the multiple element audit branch networks.

Images