Method and system for updating geospatial information based on reliability evaluation, and computer program for the same

Abstract

A method for updating spatial information based on reliability evaluation is executed by a computing device and may include: receiving image data collected from a drone or satellite; calculating an integrated reliability score by analyzing a plurality of evaluation items including two or more of the quality index, occlusion, and recency of the image data; determining an area where the integrated reliability score is greater than or equal to a preset threshold value as a valid area, and determining an area where the score is less than the threshold value as a low-reliability area; and locally updating a corresponding area of ​​an existing vector map based on the image data of the valid area.

Description

Technology Field

[0001] The embodiments relate to a method and system for updating spatial information based on reliability evaluation and a computer program for the same. More specifically, the embodiments relate to a technique for calculating an integrated reliability score by analyzing the quality index, occlusion, and recency of image data collected from a drone or satellite, and locally updating an existing vector map based on image data of a valid area where the reliability score is greater than or equal to a reference value. Background Technology

[0002] In the fields of military operations and national defense, spatial information is utilized for various purposes, such as operational planning, route finding, and facility management, and its importance is growing day by day. In particular, there is an increasing number of cases where spatial information from diverse sources, such as the Defense Topographic Database (FDB), national base maps, and civilian map services (OSM), is integrated and utilized.

[0003] As a related prior art, Registered Patent Publication No. 10-1338243 discloses a military information visualization system utilizing a web-based virtual space. The aforementioned prior art visualizes military documents in a three-dimensional virtual space and provides a zoom-in function through a Level of Detail (LOD) function, and aims to improve the method of representing military information.

[0004] Meanwhile, in wartime situations, securing up-to-date spatial information that rapidly reflects changes in terrain acts as a critical factor determining the success or failure of operations. While various technologies have been proposed to align image data collected from drones or satellites with existing maps, these conventional technologies primarily focus on image position alignment and have not sufficiently considered technologies for evaluating the quality or reliability of collected image data to utilize it selectively.

[0005] In particular, during wartime or disaster situations, it frequently happens that parts of the image are obscured by clouds, smoke, or flames, or that image quality deteriorates due to poor shooting conditions. If such low-quality images are used directly for map updates, the accuracy of spatial information may actually be compromised; however, conventional technology has not existed to systematically evaluate image reliability to determine whether or not to update the map.

[0006] Furthermore, with the recent advancement of generative AI technology making it possible to generate fake images that are difficult to distinguish from reality, serious problems could arise if AI-generated footage disguised as drone or satellite imagery is introduced into spatial information databases. However, conventional technology has not had a mechanism to verify whether collected footage is actual footage or footage generated by AI. Prior art literature

[0007] Registered Patent Publication No. 10-1338243 The problem to be solved

[0008] According to one aspect of the present invention for solving the problems of the prior art, a reliability evaluation-based spatial information update method and system and a computer program for the same can be provided, which can simultaneously ensure the accuracy and timeliness of spatial information by analyzing the quality index, occlusion, and timeliness of image data collected from a drone or satellite to calculate an integrated reliability score, and selectively utilizing only image data of valid areas where the reliability score is greater than or equal to a reference value to locally update an existing vector map. means of solving the problem

[0009] A method for updating spatial information based on reliability evaluation according to one aspect of the present invention is executed by a computing device and comprises: receiving image data collected from a drone or satellite; calculating an integrated reliability score by analyzing a plurality of evaluation items including two or more of the quality index, occlusion, and recency of the image data; determining an area where the integrated reliability score is greater than or equal to a preset threshold value as a valid area, and determining an area where the score is less than the threshold value as a low-reliability area; and locally updating a corresponding area of ​​an existing vector map based on the image data of the valid area.

[0010] In one embodiment, the step of calculating the integrated reliability score may include: collecting social data for the same location as the image data from a private data server; and cross-verifying the reliability of the image data using the social data. In this case, the cross-verification step may include applying a correction coefficient pre-set for each shooting platform.

[0011] A method for updating spatial information based on reliability evaluation according to one embodiment further comprises: a step of inferring spatial information of the low-reliability region by inputting image feature values ​​of the valid region into a pre-trained AI model for the low-reliability region; and a step of providing a user interface that induces the collection of actual measurement data and assigns a tag indicating that the inferred spatial information is inferred data to the inferred spatial information.

[0012] A reliability evaluation-based spatial information update method according to one embodiment further includes the step of comparing the image data with the image history of the same point in the past, and when a sudden change exceeding a preset threshold is detected, suspending the determination of the integrated reliability score and performing multi-source cross-verification using data collected from multiple different sources.

[0013] A spatial information update method based on reliability evaluation according to one embodiment further includes a step of verifying whether the image data is an image generated by AI. In this case, the verification step includes a step of determining whether the image is an AI-generated image by analyzing one or more of the consistency of shadow direction, geometric distortion pattern, and physical consistency of reflection pattern within the image data.

[0014] A reliability evaluation-based spatial information update system according to one embodiment includes: a memory in which executable instructions are stored; a communication unit configured to communicate with an external system; and a processor connected to the memory and the communication unit to communicate with each other.

[0015] At this time, the processor is configured to receive image data collected from a drone or satellite by executing the above command, calculate an integrated reliability score by analyzing a plurality of evaluation items including two or more of the quality index, occlusion, and recency of the image data, determine an area where the integrated reliability score is greater than or equal to a preset threshold value as a valid area, determine an area where the score is less than the threshold value as a low reliability area, and locally update the corresponding area of ​​an existing vector map based on the image data of the valid area.

[0016] A computer program according to one aspect of the present invention is stored in a computer-readable recording medium to be combined with hardware to execute a reliability evaluation-based spatial information update method according to the embodiments described above. Effects of the invention

[0017] According to a reliability evaluation-based spatial information update system and method of one aspect of the present invention, an integrated reliability score is calculated by analyzing the quality index, occlusion, and recency of image data collected from a drone or satellite, and by selectively utilizing only the image data of valid areas where the reliability score is greater than or equal to a reference value to locally update an existing vector map, the deterioration of the accuracy of spatial information caused by low-quality images can be prevented.

[0018] A reliability evaluation-based spatial information update system and method according to one aspect of the present invention improves the accuracy of reliability evaluation through cross-validation using social data and the application of correction coefficients for each shooting platform, and supplements spatial information through AI-based inference for low-reliability areas, thereby enabling the provision of continuous spatial information even in areas obscured by clouds or smoke.

[0019] Furthermore, the reliability evaluation-based spatial information update system and method according to one aspect of the present invention has the advantage of being able to prevent fake images manipulated by generative AI from entering a spatial information database in advance by analyzing the consistency of shadow directions, geometric distortion patterns, and physical consistency of reflection patterns within image data to determine whether the image is AI-generated. Brief explanation of the drawing

[0020] FIG. 1 is a conceptual diagram showing the overall concept of a reliability evaluation-based spatial information update system according to one embodiment. FIG. 2 is a schematic block diagram showing the hardware configuration of a reliability evaluation-based spatial information update system according to one embodiment. FIG. 3 is a flowchart showing each step of a reliability evaluation-based spatial information update method according to one embodiment. FIG. 4 is a flowchart illustrating the process of calculating an integrated reliability score according to one embodiment. FIGS. 5A and FIGS. 5B are conceptual diagrams illustrating examples of region-specific reliability maps according to one embodiment. FIG. 6 is a flowchart illustrating a social data-based cross-validation process according to one embodiment. FIG. 7 is a flowchart illustrating an AI-based low-trust region inference process according to one embodiment. FIGS. 8a and FIGS. 8b are exemplary diagrams showing a comparison of AI inference before and after in a low-confidence region according to one embodiment. FIG. 9 is a flowchart illustrating a multi-source cross-verification process when detecting a sudden change according to one embodiment. FIG. 10 is a flowchart illustrating an AI-generated image verification process according to one embodiment. FIGS. 11a and FIGS. 11b are conceptual diagrams illustrating examples of AI-generated image verification items according to one embodiment. Specific details for implementing the invention

[0021] In describing the embodiments of this specification, if it is determined that a detailed description of known configurations or functions could obscure the essence of the embodiments of this specification, such detailed description is omitted. Additionally, parts of the drawings unrelated to the description of the embodiments of this specification have been omitted, and similar parts are denoted by similar reference numerals.

[0022] In the embodiments of this specification, when a component is described as being "connected," "combined," or "joined" with another component, this may include not only a direct connection but also an indirect connection in which another component exists in between. Furthermore, when a component is described as "comprising" or "having" another component, this means that, unless specifically stated otherwise, it does not exclude the other component but may include additional components.

[0023] In the embodiments of this specification, terms such as first, second, etc. are used solely for the purpose of distinguishing one component from another component and do not limit the order or importance of the components unless specifically stated otherwise. Accordingly, within the scope of the embodiments of this specification, the first component in an embodiment may be referred to as the second component in another embodiment, and likewise, the second component in an embodiment may be referred to as the first component in another embodiment.

[0024] In the embodiments of this specification, distinct components are used to clearly describe their respective features and do not imply that the components are necessarily separate. The components described in the various embodiments are not necessarily essential components, and some may be optional. Accordingly, embodiments consisting of a subset of the components described in the embodiments are also included within the scope of the embodiments of this specification. Furthermore, embodiments including additional components in addition to the components described in the various embodiments are also included within the scope of the embodiments of this specification.

[0025] Hereinafter, embodiments of the present specification will be examined in detail with reference to the drawings.

[0026] FIG. 1 is a conceptual diagram showing the overall concept of a reliability evaluation-based spatial information update system according to one embodiment.

[0027] Referring to FIG. 1, a reliability evaluation-based spatial information update system according to one embodiment may be configured to receive drone images (101), satellite images (102) and / or social data (103) as inputs, and to evaluate the reliability of image data through a reliability evaluation engine (100) to update an existing vector map.

[0028] Drone imagery (101) is image data captured through a camera mounted on a drone, and is captured at a relatively low altitude, so it has high resolution but the shooting range may be limited. Satellite imagery (102) is image data captured through a sensor mounted on a satellite, and can capture a wide area, but the quality may be affected by weather conditions. Social data (103) is data collected from private data servers and may include photos uploaded to social media, review images from map services, real-time traffic information, etc., and can be used to cross-verify the reliability of drone imagery (101) or satellite imagery (102).

[0029] The reliability evaluation engine (100) receives drone images (101) and satellite images (102) and calculates an integrated reliability score by analyzing multiple evaluation items including quality index, occlusion, and recency of each image data. The reliability evaluation engine (100) determines an area where the calculated integrated reliability score is greater than or equal to a preset threshold value as a valid area (121), and determines an area where the score is less than the threshold value as a low reliability area (122).

[0030] In one embodiment, the reliability evaluation engine (100) can improve the accuracy of the reliability evaluation by cross-verifying the reliability of the image data using social data (103) and applying a correction coefficient pre-set for each shooting platform.

[0031] The valid area (121) is an area where the integrated reliability score is greater than or equal to a reference value, and the image data of the area is converted into vector data through a direct vectorization (131) process. Direct vectorization (131) refers to a process of extracting spatial objects such as roads, buildings, and terrain from image data and converting them into a vector format.

[0032] The low-reliability area (122) is an area where the integrated reliability score is below the reference value, and corresponds to an area where quality is degraded due to being obscured by clouds, smoke, flames, etc. or poor shooting conditions. For the low-reliability area (122), spatial information is supplemented through an AI inference (132) process.

[0033] AI inference (132) refers to the process of inferring spatial information of a low-reliability area (122) by inputting image feature values ​​of a valid area (121) and existing vector map data into a pre-trained AI model. Spatial information generated through AI inference (132) is assigned a tag indicating that it is inference data, and a user interface that induces the collection of actual measurement data may be provided.

[0034] Local map updating (140) refers to the process of reflecting vector data generated through direct vectorization (131) and AI inference (132) to the corresponding area of ​​an existing vector map. Local map updating (140) does not update the entire map, but rather performs partial updating limited to the area where new image data is collected, thereby improving updating efficiency and preventing unnecessary data changes.

[0035] The integrated vector map (150) is a vector map that reflects the latest spatial information through local map updates (140), and includes accurate spatial objects extracted from valid areas (121) and spatial objects supplemented through AI inference (132) in low-reliability areas (122).

[0036] In this way, a reliability evaluation-based spatial information update system according to one embodiment can simultaneously ensure the accuracy and up-to-dateness of spatial information by systematically evaluating the reliability of image data collected from a drone or satellite and selectively utilizing only the data of a valid area with high reliability to locally update an existing vector map.

[0037] FIG. 2 is a schematic block diagram showing the hardware configuration of a reliability evaluation-based spatial information update system according to one embodiment.

[0038] The reliability evaluation-based spatial information update system according to the embodiments is implemented in the form of any computing device and may be entirely hardware or have aspects that are partially hardware and partially software.

[0039] For example, the system, device, server, and each unit included therein according to the embodiments may collectively refer to a device for exchanging data of a specific format and content via electronic communication and related software. In this specification, terms such as “unit,” “module,” “server,” “system,” “platform,” “device,” or “terminal” are intended to refer to a combination of hardware and software driven by said hardware. For example, the hardware may be a data processing device including a CPU or other processor. Additionally, the software driven by the hardware may refer to a running process, object, executable, thread of execution, program, etc.

[0040] A reliability evaluation-based spatial information update system (2) according to embodiments may be implemented as a computing device comprising one or more hardware components. In this case, the reliability evaluation-based spatial information update system (2) may include a memory (210), a processor (220), a communication unit (230), a spatial information database (DB) (240), and an AI inference module (250). In one embodiment, the reliability evaluation-based spatial information update system (2) may further include an input / output unit (not shown).

[0041] The reliability evaluation-based spatial information update system (2) according to the embodiments may be implemented in the form of any computing device, such as a server, a mobile communication terminal, a notebook computer, a personal computer, a PDA (personal digital assistant), a tablet, or a set-top box for IPTV (Internet Protocol Television).

[0042] The memory (210) is a non-transient computer-readable recording medium and may include a permanent mass storage device such as RAM (random access memory), ROM (read only memory), disk drive, SSD (solid state drive), flash memory, etc. Here, the permanent mass storage device such as ROM, SSD, flash memory, disk drive, etc. may be included in the device or server described above as a separate permanent storage device distinct from the memory (210).

[0043] Additionally, the memory (210) may store an operating system and at least one program code (e.g., code for executing an application installed and running on a server or user device). These software components may be loaded from a computer-readable recording medium separate from the memory (210). This separate computer-readable recording medium may include computer-readable recording media such as a floppy drive, disk, tape, DVD / CD-ROM drive, or memory card.

[0044] In another embodiment, software components may be loaded into memory (210) via a communication unit (230) rather than a computer-readable recording medium. For example, at least one program may be loaded into memory (210) based on a computer program installed by files provided over a network by developers or a file distribution system (e.g., a commercial application store service server) that distributes installation files for applications.

[0045] The processor (220) may be configured to process instructions of a computer program by performing basic arithmetic, logic, and input / output operations. Instructions may be provided to the processor (220) by memory (210) or a communication unit (230). For example, the processor (220) may be configured to execute instructions received according to program code stored in a recording device such as memory (210).

[0046] The communication unit (230) can provide a function for the reliability evaluation-based spatial information update system (2) to collect drone images (111), satellite images (112), and social data (113) through the network (132). The drone images (111) are image data captured through a camera mounted on a drone, the satellite images (112) are image data captured through a sensor mounted on a satellite, and the social data (113) may include SNS photos, map service review images, etc., collected from a private data server. Additionally, the function of the communication unit (230) can be controlled by a processor (220) that references the memory (210).

[0047] The spatial information DB (240) is a database that stores various spatial objects, such as roads, buildings, and terrain, in the form of vector data. Spatial information collected from various sources, such as the Defense Terrain Database (FDB), national base maps, and civilian map services (OSM), can be integrated and stored in the spatial information DB (240). The spatial information DB (240) is connected to communicate with the processor (220) and can provide existing vector map data and store updated spatial objects in response to a request from the processor (220).

[0048] The AI ​​inference module (250) is a component for performing spatial information inference for a low-reliability area. The AI ​​inference module (250) may include a pre-trained AI model that infers spatial information of a low-reliability area by taking image feature values ​​of a valid area and existing vector map data as input. The AI ​​inference module (250) may be implemented by including a computational accelerator such as a GPU (Graphics Processing Unit) or an NPU (Neural Processing Unit) and may operate under the control of the processor (220).

[0049] The input / output unit (not shown) may be a means for interfacing with an external input / output device (not shown). For example, the external input device may include devices such as a keyboard, mouse, microphone, or camera, and the external output device may include devices such as a display, speaker, or haptic feedback device. As another example, the input / output unit may be a means for interfacing with a device in which the functions for input and output are integrated into one, such as a touchscreen.

[0050] In addition, in other embodiments, the reliability evaluation-based spatial information update system (2) may include more components than those shown in FIG. 2 depending on the nature of the device to which it is applied. For example, it may include other components such as a transceiver, a GPS (Global Positioning System) module, a camera, various sensors, a database, etc.

[0051] The processor (220) is configured to execute an operation corresponding to each step of the reliability evaluation-based spatial information update method according to embodiments of the present invention by executing instructions stored in memory (210).

[0052] That is, the processor (220) receives image data collected from a drone or satellite and can calculate an integrated reliability score by analyzing multiple evaluation items among the quality index, occlusion, and recency of the image data. Additionally, the processor (220) can determine an area where the integrated reliability score is greater than or equal to a preset threshold value as a valid area and an area where it is less than the threshold value as a low reliability area, and can locally update the corresponding area of ​​an existing vector map stored in the spatial information DB (240) based on the image data of the valid area.

[0053] A method for updating spatial information based on reliability evaluation according to embodiments executed by a processor (220) is described in more detail below.

[0054] FIG. 3 is a flowchart showing each step of a reliability evaluation-based spatial information update method according to one embodiment.

[0055] Referring to FIG. 3, a reliability evaluation-based spatial information update method according to one embodiment may be executed by a computing device, and the computing device may be a reliability evaluation-based spatial information update system (2) described with reference to FIG. 2.

[0056] First, the processor (220) can receive image data collected from a drone or satellite (S310). The image data can be received in the form of drone image (111) or satellite image (112) via the network (132) through the communication unit (230). The image data may include metadata such as shooting location, shooting time, shooting altitude, and sensor information.

[0057] Next, the processor (220) can analyze two or more information items among the quality index, occlusion, and recency of the received image data, or all of them (S320). The quality index is an indicator that comprehensively evaluates the resolution, clarity, noise level, etc. of the image, the occlusion indicates the percentage of the surface of the earth that is obscured by clouds, smoke, flames, etc., and recency is an indicator that reflects the time difference between the time the image was taken and the current time. The processor (220) can calculate individual scores for each evaluation item.

[0058] Next, the processor (220) can calculate an integrated reliability score based on the analyzed evaluation items (S330). The integrated reliability score can be calculated by weighting the individual scores for quality index, occlusion, and recency, and can have a value between 0.0 and 1.0. At this stage, the processor (220) can correct for characteristic differences between the drone and the satellite by applying a pre-set correction coefficient for each shooting platform. Additionally, the processor (220) can improve the accuracy of the integrated reliability score by cross-verifying the reliability of the image data using social data (113).

[0059] Next, the processor (220) can compare the calculated integrated reliability score with a preset reference value (S340). The reference value may be set differently depending on the purpose of use of the image data, the required level of accuracy, etc., and may be set to a value such as 0.7 or 0.8.

[0060] In one embodiment, the process of steps S320 to S340 can be performed integrally by a pre-trained machine learning model. Specifically, the processor (220) inputs received image data and metadata included in the image data (shooting location, shooting time, shooting altitude, sensor information, etc.) into the machine learning model, and the machine learning model automatically extracts features regarding quality index, occlusion, and recency from the input image data to calculate an integrated reliability score, and can determine whether each area of ​​the image data corresponds to a valid area or a low-reliability area based on the calculated integrated reliability score.

[0061] In this case, the machine learning model can perform the calculation of individual scores for quality index, occlusion, and recency, the application of correction coefficients per shooting platform, the calculation of an integrated confidence score through the weighted average of individual scores, and the determination of valid and low-confidence areas through comparison with reference values ​​in an end-to-end manner within a single model.

[0062] In this case, the machine learning model can be pre-trained using previously collected image data, information on valid and low-confidence regions labeled by experts on the image data, and the actual utilization results of the image data (success of map updates, accuracy of the updated map, etc.). Accordingly, even if the scoring criteria and weights for individual evaluation items, as well as the threshold values ​​for determining valid regions, are not explicitly set, the machine learning model can automatically learn and apply optimal judgment criteria from the training data.

[0063] In one embodiment, the machine learning model may receive the purpose of use of the image data (precision map creation, disaster response, operational planning, etc.) and the characteristics of the target area (urban area, mountainous area, coastal area, etc.) as additional inputs, and adaptively apply different judgment criteria according to the purpose of use and regional characteristics.

[0064] As a result of comparison, for areas where the integrated reliability score is greater than or equal to the reference value, the processor (220) can determine that area as a valid area (S350). A valid area refers to an area where the quality of the image data is sufficiently high to be directly used for map updating.

[0065] Meanwhile, for areas where the integrated reliability score is below the reference value, the processor (220) may determine that the area is a low-reliability area (S360). A low-reliability area is an area that is obscured by clouds, smoke, etc., or has degraded quality due to poor shooting conditions, and is difficult to use for direct map updates.

[0066] Finally, the processor (220) can locally update the corresponding area of ​​the existing vector map stored in the spatial information DB (240) based on the image data of the valid area (S370). Local map updating means that the entire map is not updated, but rather the update is performed partially, limited to the area where new image data has been collected. For the valid area, spatial objects extracted from the image data are reflected in the existing vector map, and for the low-reliability area, spatial information inferred through the AI ​​inference module (250) can be reflected.

[0067] In this way, the reliability evaluation-based spatial information update method according to one embodiment systematically evaluates the reliability of image data and selectively utilizes only the data of valid areas with high reliability to locally update the existing vector map, thereby preventing the deterioration of the accuracy of spatial information caused by low-quality images.

[0068] In one embodiment, when multiple image data exist for the same point, the processor (220) can fuse them to obtain a higher reliability than a single image data. Specifically, when both drone image (111) and satellite image (112) are collected for the same point, the processor (220) can individually calculate an integrated reliability score for each image data and identify the valid area of ​​each image data. For example, there may be an area that is determined to be valid in the drone image (111) but determined to be a low-reliability area in the satellite image (112), and vice versa.

[0069] At this time, the processor (220) can combine the valid areas of each image data in a mutually complementary manner to minimize the area of ​​the low-reliability region compared to the case where only a single image data is used. For example, if an area obscured by clouds in the satellite image (112) is captured normally in the drone image (111), the map update can be performed for that area by utilizing the data from the drone image (111). In addition, the processor (220) can automatically detect inconsistency areas where different spatial object information is extracted for the same point from multiple image data, and designate the detected inconsistency areas as targets requiring additional verification to provide a notification to the administrator or request additional image collection.

[0070] In addition, in one embodiment, when the processor (220) performs a map update for a valid area in step S370, it may process the valid areas by assigning update priorities differentially rather than treating all valid areas equally. Specifically, the processor (220) may determine the update priority by analyzing the strategic importance, changeability, and update history for each valid area.

[0071] Strategic importance can be calculated based on whether the area corresponds to a military critical zone, a border area, or an area near major infrastructure, and high update priorities may be assigned to areas with high strategic importance. Change potential can be calculated based on whether recent rapid changes have been detected in the area, or whether the area corresponds to a planned development area or a disaster zone, and high update priorities may be assigned to areas with a high probability of change. Update history can be calculated based on the time elapsed since the last update for the area, and the update priority may be automatically raised for areas with long update cycles.

[0072] The processor (220) may perform map updates sequentially starting from high-priority areas according to the calculated update priority, or selectively perform map updates only on high-priority areas when system resources are limited.

[0073] Furthermore, in one embodiment, the processor (220) may perform automatic verification of the update result after the map update is completed in step S370 and feed the verification result back to the training of the machine learning model. Specifically, the processor (220) may automatically verify whether the updated spatial object maintains logical consistency with adjacent spatial objects of the existing vector map.

[0074] For example, it can be verified whether the updated road object is properly connected to the existing road network, whether the updated building object is located within the existing parcel boundaries, or whether the size and shape of the updated space object match the general patterns of the area.

[0075] Additionally, the processor (220) can compare and verify the accuracy between the updated spatial object and the actual measurement data or additional image data when actual measurement data or additional image data is subsequently collected for the updated area. The processor (220) stores the results of automatic verification and accuracy comparison verification in conjunction with the integrated reliability score and valid area determination results of the image data used for the update, and can use the stored verification results for retraining the machine learning model. Accordingly, the machine learning model can continuously improve the accuracy of the integrated reliability score calculation and valid area determination by reflecting the accuracy of the actual update results.

[0076] FIG. 4 is a flowchart illustrating the process of calculating an integrated reliability score according to one embodiment.

[0077] Referring to FIG. 4, the process for calculating an integrated reliability score according to one embodiment may correspond to a detailed example of the process of analyzing image data and calculating an integrated reliability score in steps S320 and S330 of FIG. 3.

[0078] First, the processor (220) may perform any two or more combinations of quality index calculation (S411), occlusion analysis (S412), and recency evaluation (S413) or all of these in parallel on the received image data.

[0079] To explain each step in detail, in the quality index calculation (S411) step, the processor (220) can calculate a quality index by analyzing the resolution, clarity, noise level, and degree of color distortion of the image data. The quality index is an indicator that quantifies the overall quality level of the image into a value between 0.0 and 1.0.

[0080] In the occlusion analysis (S412) step, the processor (220) can analyze the proportion of the area where the surface of the earth is obscured by clouds, smoke, flames, fog, etc. within the image data. The occlusion represents the proportion of the obscured area within the image, and a higher score may be assigned as the occlusion is lower (i.e., the less the obscured area).

[0081] In the recency evaluation (S413) step, the processor (220) can calculate a recency score by analyzing the time difference between the time of shooting of the image data and the current time. The recency score has a higher value the more recent the time of shooting, and can be set to gradually decrease as time passes.

[0082] In one embodiment, the processor (220) may apply a correction factor specific to the shooting platform to individual scores for quality index, occlusion, and recency (S420). The correction factor specific to the shooting platform is intended to correct for differences in shooting characteristics between drones and satellites; for example, drone images have high resolution due to low-altitude shooting but have a limited shooting range, while satellite images can capture a wide area but are affected by atmospheric conditions. The processor (220) may correct the individual scores by reflecting these platform-specific characteristics.

[0083] In one embodiment, the correction factor for each shooting platform can be dynamically adjusted according to environmental conditions at the time of shooting. Specifically, the processor (220) checks weather information (weather, wind speed, humidity, etc.), lighting conditions (solar elevation angle, time of shooting, etc.), and atmospheric conditions (fine dust concentration, visibility, etc.) at the time of shooting included in the metadata of the image data, and can dynamically adjust the correction factor according to these environmental conditions. For example, in the case of satellite imagery, if the image is shot under conditions where the concentration of fine dust in the atmosphere is high, the correction factor for the quality index can be lowered, and in the case of drone imagery, if the image is shot under strong wind conditions, the correction factor for the quality index can be lowered to reflect the decrease in shooting stability.

[0084] In addition, in one embodiment, the correction coefficient for each shooting platform can be automatically learned and updated based on past shooting history and actual usage results. Specifically, the processor (220) can receive feedback on the accuracy verification results of the updated map after video data collected from a specific shooting platform is used for map updating. Based on the received feedback, the processor (220) can automatically adjust the correction coefficient for the corresponding shooting platform so that the correction coefficient is continuously optimized to reflect the actual usage results. For example, if video data collected from a specific drone model consistently shows map update results with high accuracy, the correction coefficient for that drone model can be automatically adjusted upward.

[0085] Next, the processor (220) can calculate an integrated score by weighting the individual corrected scores (S430). When calculating the integrated score, the weights for each evaluation item may be set differently depending on the purpose of use of the image data. For example, in a disaster situation, a high weight may be given to recency, and when creating a high-precision map, a high weight may be given to the quality index.

[0086] Finally, the processor (220) can output the calculated integrated score as a final score (S440). The final score can be normalized to have a value between 0.0 and 1.0, for example, and this score is compared with a reference value in step S340 of FIG. 3 and used to determine a valid area or a low-confidence area.

[0087] FIGS. 5A and FIGS. 5B are conceptual diagrams illustrating examples of region-specific reliability maps according to one embodiment.

[0088] FIG. 5a illustrates an example of original image data received from a drone or satellite. Referring to FIG. 5a, the original image data includes a cloud area (510), a smoke area (520), and a normal shooting area (530). The cloud area (510) is located at the top left of the image, and the surface of the earth is obscured by clouds, making it difficult to directly extract spatial information of that area. The smoke area (520) is located at the top center of the image, and the surface of the earth is partially obscured by smoke caused by fire or explosion. The normal shooting area (530) is an area where the surface of the earth is clearly captured without obscuring elements such as clouds or smoke, and spatial objects such as roads, buildings, and terrain can be identified.

[0089] Figure 5b shows an example of a confidence map generated as a result of applying the integrated confidence score calculation process described with reference to Figure 4 to the original image data of Figure 5a. Referring to Figure 5b, the confidence map divides the original image data into a grid shape and displays the integrated confidence score calculated for each grid cell. A confidence score between 0.0 and 1.0 is displayed in each grid cell.

[0090] As illustrated in FIG. 5b, the grid cells corresponding to the normal shooting area (530) of FIG. 5a have a high confidence score of 0.9, which indicates that the image quality of the area is excellent and there are no occlusion elements, so it can be directly used for map updates. On the other hand, the grid cells corresponding to the cloud area (510) and smoke area (520) of FIG. 5a may have a relatively low confidence score, which indicates that the image data of the area is of degraded quality due to occlusion elements.

[0091] A legend is displayed at the bottom of FIG. 5b, indicating that areas with a confidence score of 0.8 or higher are determined to be valid areas, and areas with a confidence score of less than 0.5 are determined to be low-confidence areas. Image data of grid cells determined to be valid areas is used to update existing vector maps through a direct vectorization process, and for grid cells determined to be low-confidence areas, spatial information can be inferred through an AI inference module (250).

[0092] FIG. 6 is a flowchart illustrating a social data-based cross-validation process according to one embodiment.

[0093] Referring to FIG. 6, a social data-based cross-validation process according to one embodiment may be additionally performed to improve the accuracy of the reliability evaluation in the integrated reliability score calculation process of FIG. 4.

[0094] First, the processor (220) can extract video shooting location and time information from the received video data (S610). The video shooting location can be extracted from GPS coordinates or metadata included in the video data, and the shooting time can be extracted from the timestamp information of the video data.

[0095] Next, the processor (220) can request data for the same location from a private data server through the communication unit (230) (S620). The private data server may include a social network service (SNS) server, a map service server, a real-time traffic information server, etc., and the processor (220) can request data corresponding to the location and time period based on the extracted shooting location and time information.

[0096] Next, the processor (220) can collect social data from a private data server (S630). The social data may include photos uploaded to social media, review images from map services, real-time traffic camera footage, street view images taken by users, etc. The social data may include images taken at different points in time or at different angles for the same location, and can be used to verify the reliability of the original image data.

[0097] Next, the processor (220) can compare and verify the collected social data with the original image data (S640). In the comparison and verification step, the processor (220) can analyze whether spatial objects (roads, buildings, terrain, etc.) displayed in the original image data are confirmed identically in the social data. For example, if a specific building is displayed as existing in the original image but is confirmed to have been demolished in the social data, it can be determined that there is a problem with the timeliness of the original image. Additionally, if the shape of the road in the original image is displayed differently from the social data, it can be determined that there is a problem with the accuracy of the original image.

[0098] Finally, the processor (220) can correct the reliability score based on the comparison verification results (S650). If the accuracy of the original image data is confirmed as a result of the comparison verification with social data, the reliability score may be corrected upward, and if an inconsistency is found, the reliability score may be corrected downward. The corrected reliability score is compared with a reference value in step S340 of FIG. 3 and used to determine a valid area or a low-reliability area.

[0099] FIG. 7 is a flowchart illustrating an AI-based low-trust region inference process according to one embodiment.

[0100] Referring to FIG. 7, an AI-based low-reliability region inference process according to one embodiment can be performed to supplement spatial information for a region determined as a low-reliability region in step S360 of FIG. 3.

[0101] First, the processor (220) can identify a low-reliability area determined in step S360 of FIG. 3 (S710). A low-reliability area is an area where the integrated reliability score is less than a preset threshold value, and corresponds to an area where quality is degraded due to being obscured by clouds, smoke, flames, etc., or due to poor shooting conditions.

[0102] Next, the processor (220) can extract image feature values ​​from image data of the valid area (S720). The image feature values ​​may include spatial characteristics such as the width of the road, the shape of the building, the slope of the terrain, and the type of land cover identified in the valid area. Additionally, the image feature values ​​may include information such as spatial object connection patterns, directionality, and continuity of the valid area adjacent to the low-reliability area.

[0103] Next, the processor (220) can load existing vector map data corresponding to a low-reliability area from the spatial information DB (240) (S730). The existing vector map data is spatial information of the area constructed at a past point in time and may include road networks, building layouts, topographic information, etc. The existing vector map data is used as reference information for inferring spatial information of the low-reliability area.

[0104] Next, the processor (220) can input the extracted image feature values ​​and existing vector map data into the AI ​​model of the AI ​​inference module (250) (S740). The AI ​​model may be a deep learning model that has been pre-trained to infer spatial information of the low-reliability area based on the image feature values ​​of the valid area and the existing vector map data. The AI ​​model can, for example, infer the extension line of a road connecting the valid area to the low-reliability area, or infer whether a building that existed in the existing vector map still exists.

[0105] Next, the processor (220) can infer spatial information of a low-reliability area based on the output of the AI ​​model (S750). The inferred spatial information includes information about spatial objects such as roads, buildings, and terrain within the low-reliability area, and each spatial object may be assigned an inference reliability score.

[0106] Finally, the processor (220) can output inferred spatial information (S760). At this stage, the inferred spatial information is assigned a tag indicating that it is inferred data, so that it can be distinguished from spatial information extracted from actual shooting data. Additionally, the processor (220) can provide a user interface that induces the collection of actual measurement data for the inferred spatial information, and the inferred spatial information can be automatically replaced with actual measurement data when valid image data for the area is collected thereafter.

[0107] FIGS. 8a and FIGS. 8b are exemplary diagrams showing a comparison of AI inference before and after in a low-confidence region according to one embodiment.

[0108] FIG. 8a shows the image data and vector map state before AI inference is performed. Referring to FIG. 8a, the image data includes a cloud area (810) and a smoke area (820), so some spatial objects are obscured. The cloud area (810) is located at the top and bottom left of the image, and in that area, spatial objects such as roads and buildings are partially obscured. The smoke area (820) is located at the center and right of the image, and the road is partially obscured by smoke caused by a fire or explosion.

[0109] In FIG. 8a, roads are indicated by diagonal lines (diagonal pattern), and the road connections are shown as broken in sections obscured by cloud areas (810) and smoke areas (820). Buildings are indicated by a grid pattern, and some buildings are obscured by cloud areas (810). As such, in low-reliability areas, spatial objects cannot be directly extracted from image data, resulting in gaps in the vector map.

[0110] FIG. 8b shows the state of the vector map after the AI-based low-reliability area inference process described with reference to FIG. 7 has been performed. Referring to FIG. 8b, it can be seen that the road sections that were obscured by the cloud area (810) and smoke area (820) in FIG. 8a are connected through AI inference. The AI ​​inference module (250) inferred and connected the road paths within the low-reliability area based on feature values ​​such as the direction, width, and curvature of the road extracted from the valid area and existing vector map data.

[0111] In addition, the building area that was obscured by the cloud area (810) in FIG. 8a is also displayed in its complete form in FIG. 8b. The AI ​​inference module (250) inferred the existence and shape of the obscured building based on the shape of the building existing in the existing vector map and the building placement pattern of the adjacent valid area.

[0112] The spatial object inferred in Fig. 8b is displayed in the same form as the spatial object extracted from actual shooting data to form a continuous vector map, but internally it is tagged to indicate that it is inferred data, so that when valid image data for the area is collected later, it can be automatically replaced with actual data.

[0113] FIG. 9 is a flowchart illustrating a multi-source cross-verification process when detecting a sudden change according to one embodiment.

[0114] Referring to FIG. 9, the multi-source cross-verification process when detecting a sudden change according to one embodiment can be additionally performed when a sudden change in image data is detected in the reliability evaluation-based spatial information update method of FIG. 3.

[0115] First, the processor (220) can load past image history for the same point as the currently received image data from the spatial information DB (240) (S910). The past image history may include drone images, satellite images, or spatial object information extracted therefrom that were previously collected for the point.

[0116] Next, the processor (220) can compare the loaded past image history with the currently received image data (S920). The comparison items may include the presence and shape of roads, the presence and arrangement of buildings, changes in terrain, changes in land cover, etc. The processor (220) can calculate the amount of change by quantifying the difference between the past image and the current image.

[0117] Next, the processor (220) can compare the calculated amount of change with a preset threshold value (S930). The threshold value is a reference value for distinguishing between normal change and sudden change, and can be individually set for, for example, building loss rate, road deformation rate, terrain change rate, etc.

[0118] As a result of comparison, if the amount of change is less than the threshold, the processor (220) can proceed with normal processing (S940). Normal processing refers to the process of calculating the integrated reliability score and determining the valid area / low reliability area as described with reference to FIG. 3, and a general map update procedure is performed by determining that the change in image data is within the normal range.

[0119] Meanwhile, if the amount of change exceeds a threshold value, the processor (220) may withhold the determination of the integrated reliability score (S950). The sudden change may be due to actual terrain changes (e.g., disaster, development, etc.), but may also be due to errors in image data, fake images manipulated by AI, or abnormalities in shooting conditions, so additional verification is required.

[0120] Next, the processor (220) can collect data about the point from multiple different sources through the communication unit (230) (S960). The multi-source data may include drone or satellite images taken at different times, social data (SNS photos, map review images, etc.), news reports, information from disaster management systems, etc.

[0121] Finally, the processor (220) can perform cross-verification using the collected multi-source data (S970). As a result of the cross-verification, if the same change is confirmed in multiple sources, it is determined to be an actual terrain change and the map update is performed; if the change appears only in the current image data and is not confirmed in other sources, it is determined to be an error or manipulation of the image data and the corresponding image data is excluded from the map update.

[0122] FIG. 10 is a flowchart illustrating an AI-generated image verification process according to one embodiment.

[0123] Referring to FIG. 10, the AI-generated image verification process according to one embodiment is performed before evaluating the reliability of the image data in the reliability evaluation-based spatial information update method of FIG. 3, thereby preventing fake images manipulated by generative AI from entering the spatial information database in advance. Additionally, the AI-generated image verification process according to this embodiment may be applied to verify whether the image included in the social data is not a model generated by an AI model in the comparison verification process (S640) between the social data and the original image shown in FIG. 4.

[0124] First, the processor (220) can receive image data collected from a drone or satellite as input data (S1010).

[0125] Next, the processor (220) can perform multiple physical consistency analyses on the received image data in parallel. Specifically, the processor (220) can perform shadow direction analysis (S1021), geometric distortion analysis (S1022), reflection pattern analysis (S1023) and / or noise pattern analysis (S1024).

[0126] To explain each process in detail, in the shadow direction analysis (S1021), the processor (220) can analyze whether the direction of shadows generated from multiple objects in the image is consistent. In actual captured images, the shadows of all objects captured at the same time have a consistent direction according to the position of the sun, but in images generated by AI, there may be cases where the shadow directions are inconsistent.

[0127] In the geometric distortion analysis (S1022), the processor (220) can analyze whether the geometric distortion pattern appearing in spatial objects such as buildings and roads within the image matches the characteristics of the shooting platform. In drone or satellite imagery, characteristic geometric distortion patterns appear depending on the shooting altitude, lens characteristics, shooting angle, etc., but in images generated by AI, such patterns may be unnatural or absent.

[0128] In the reflection pattern analysis (S1023), the processor (220) can analyze whether the reflection pattern appearing on the water surface, glass window, metal surface, etc., in the image has physical consistency. In actual captured images, a reflection pattern that conforms to physical laws appears depending on the position of the light source and the characteristics of the reflective surface, but in images generated by AI, the reflection pattern may appear unrealistically.

[0129] In the noise pattern analysis (S1024), the processor (220) can analyze whether the distribution pattern of noise in the image matches the characteristics of the actual camera sensor. In the actual captured image, a unique noise pattern according to the sensor characteristics appears, but in the image generated by AI, the noise pattern may appear artificial or uniform.

[0130] Next, the processor (220) can determine the physical consistency by synthesizing the plurality of analysis results (S1030). The processor (220) can calculate a consistency score for each analysis item and calculate a weighted average of them to determine the overall physical consistency score.

[0131] Next, the processor (220) can determine whether the image is fake based on the overall physical consistency score (S1040). The processor (220) can determine the image as normal if the overall physical consistency score is greater than or equal to a preset threshold value, and determine the image as suspected of being fake if it is less than the threshold value.

[0132] If the result of the judgment is determined to be a normal image, the processor (220) can perform a reliability evaluation as described with reference to FIG. 3 (S1051).

[0133] Meanwhile, if it is determined that a false image is suspected, the processor (220) may perform false image response measures (S1052). False image response measures may include excluding the image data in question from map updates, sending a warning notification to an administrator, and recording the source of the image data in question for future analysis.

[0134] In this way, the AI-generated image verification process according to one embodiment can detect fake images generated by AI by analyzing the physical consistency of shadow direction, geometric distortion, reflection pattern, and noise pattern, thereby preventing manipulated images from entering the spatial information database in advance.

[0135] FIGS. 11a and FIGS. 11b are conceptual diagrams illustrating examples of AI-generated image verification items according to one embodiment.

[0136] FIGS. 11a and FIGS. 11b show specific examples of shadow direction consistency verification performed in the shadow direction analysis (S1021) of FIG. 10.

[0137] FIG. 11a shows an example of a false image generated by AI. Referring to FIG. 11a, there are two buildings in the image, and the sun is located in the upper left corner. The shadow (1110) of the first building is formed in the lower right direction, while the shadow (1120) of the second building is formed in the lower left direction. In an image taken at the same time and at the same sun position, the shadows of all buildings should be formed in the same direction, but in FIG. 11a, the shadow directions of the two buildings appear different from each other. This discrepancy in shadow direction is a characteristic of physical inconsistency that commonly occurs in images generated by AI, and can be detected through the shadow direction analysis (S1021) of FIG. 10.

[0138] FIG. 11b shows an example of a normal image captured in reality. Referring to FIG. 11b, there are two buildings in the image, and the sun is located in the upper left corner. The shadow (1130) of the first building and the shadow (1140) of the second building are both consistently formed in the direction of the lower right corner. This is the direction of the shadow that is naturally formed according to the laws of physics when the sun is located in the upper left corner, and corresponds to the characteristics of the actual captured image.

[0139] In the shadow direction analysis (S1021) of FIG. 10, the processor (220) can extract direction vectors of shadows generated from multiple objects in the image and calculate the deviation between the extracted direction vectors to calculate a shadow direction consistency score. When the shadow direction is inconsistent as in FIG. 11a, a low consistency score is calculated and the image is judged as suspected of being a false image, and when the shadow direction is consistent as in FIG. 11b, a high consistency score is calculated and the image is judged as normal.

[0140] The operation of the reliability evaluation-based spatial information update method according to the embodiments described above may be implemented at least partially as a computer program and recorded on a computer-readable recording medium. A computer-readable recording medium on which a program for implementing the operation according to the methods of the embodiments is recorded includes all types of recording devices in which data readable by a computer is stored. Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage device, etc. Additionally, the computer-readable recording medium may be distributed across networked computer systems, and computer-readable code may be stored and executed in a distributed manner. Furthermore, the functional programs, codes, and code segments for implementing the present embodiment will be easily understood by a person skilled in the art to which the present embodiment belongs.

[0141] Additionally, each block or each step illustrated in the flowcharts of this specification may represent a module, segment, or part of code comprising one or more executable instructions for executing a specified logical function(s). Furthermore, in some alternative embodiments, the functions mentioned in the blocks or steps may occur out of order. For example, two blocks or steps illustrated in succession may actually be performed substantially simultaneously, or the blocks or steps may occasionally be performed in reverse order according to the corresponding function.

[0142] The present invention described above has been explained with reference to the embodiments illustrated in the drawings, but this is merely illustrative and those skilled in the art will understand that various modifications and variations of the embodiments are possible therefrom. However, such modifications should be considered to be within the technical scope of protection of the present invention. Accordingly, the true technical scope of protection of the present invention should be determined by the technical spirit of the appended claims.

Claims

Claim 1 A method for updating spatial information based on reliability evaluation executed by a computing device, comprising: receiving image data collected from a drone and a satellite; calculating an integrated reliability score by analyzing a plurality of evaluation items including two or more of the quality index, occlusion, and recency of the image data using a first model, which is a pre-trained machine learning model; determining an area where the integrated reliability score is greater than or equal to a preset threshold value as a valid area and determining an area where the integrated reliability score is less than the threshold value as a low-reliability area using the first model; locally updating a corresponding area of ​​an existing vector map based on the image data of the valid area; inferring spatial information of the low-reliability area by inputting the image feature values ​​of the valid area into a second model, which is a pre-trained AI model, for the low-reliability area; assigning a tag indicating that the inferred spatial information is inferred data to the inferred spatial information and providing a user interface that induces the collection of actual measurement data; and receiving the actual measurement data for the low-reliability area. and, in response to the reception of the actual measurement data, the method includes the step of replacing the inferred spatial information with the actual measurement data; the step of receiving the image data includes the step of receiving the first image data collected from a drone and the second image data collected from a satellite, respectively, corresponding to the same point; the step of calculating the integrated reliability score includes the step of calculating the integrated reliability score for each of the first image data and the second image data; the step of determining includes the step of identifying the valid area from each of the first image data and the second image data based on the integrated reliability score; and the step of locally updating includes the step of combining the valid area of ​​the first image data and the valid area of ​​the second image data so as to minimize the area of ​​the low reliability region.A method for updating spatial information based on reliability evaluation, comprising: a step of determining an update priority by analyzing one or more of strategic importance, changeability, and update history for the valid area; and a step of selectively updating the valid area based on the update priority; a step of verifying whether spatial objects within the updated valid area maintain logical consistency with spatial objects of the existing vector map after the update of the valid area is completed; a step of storing the result of the verification step in conjunction with the integrated reliability score and the determination result of the valid area; and a step of retraining the first model using the data stored in the storage step. Claim 2 A method for updating spatial information based on reliability evaluation according to claim 1, wherein the step of calculating the integrated reliability score comprises: a step of collecting social data for the same point as the image data from a private data server; and a step of cross-verifying the reliability of the image data using the social data, wherein the cross-verification step comprises a step of applying a correction coefficient pre-set for each shooting platform. Claim 3 delete Claim 4 A method for updating spatial information based on reliability evaluation according to claim 1, further comprising the step of, when a sudden change exceeding a preset threshold is detected by comparing the image data with the image history of the same point in the past, suspending the determination of the integrated reliability score and performing multi-source cross-verification using data collected from multiple different sources. Claim 5 A method for updating spatial information based on reliability evaluation according to claim 1, further comprising a step of verifying whether the image data is an image generated by AI, wherein the verification step comprises a step of determining whether the image is an AI-generated image by analyzing one or more of the consistency of shadow direction, geometric distortion pattern, and physical consistency of reflection pattern within the image data. Claim 6 A computer program stored on a computer-readable recording medium to execute a reliability evaluation-based spatial information update method according to any one of paragraphs 1, 2, 4, and 5 in combination with hardware. Claim 7 Memory storing executable commands; a communication unit configured to communicate with an external system; A reliability evaluation-based spatial information update system comprising a processor connected to communicate with the memory and the communication unit, wherein the processor receives image data collected from a drone and a satellite by executing the command, calculates an integrated reliability score by analyzing a plurality of evaluation items including two or more of the quality index, occlusion, and recency of the image data using a first model which is a pre-trained machine learning model, determines an area where the integrated reliability score is greater than or equal to a preset threshold value as a valid area, and determines an area where the integrated reliability score is less than the threshold value as a low-reliability area, locally updates the corresponding area of ​​an existing vector map based on the image data of the valid area, infers spatial information of the low-reliability area by inputting the image feature values ​​of the valid area into a second model which is a pre-trained AI model for the low-reliability area, assigns a tag indicating that the inferred spatial information is inferred data to the inferred spatial information, and provides a user interface that induces the collection of actual measurement data, receives the actual measurement data for the low-reliability area, and in response to the receipt of the actual measurement data, replaces the inferred spatial information with the actual measurement data, wherein at the same location Accordingly, first image data collected from a drone and second image data collected from a satellite are each received, and the integrated reliability score is calculated for each of the first image data and the second image data; based on the integrated reliability score, the valid area is identified from each of the first image data and the second image data, and in locally updating the corresponding area of ​​an existing vector map based on the image data of the valid area, the valid area of ​​the first image data and the valid area of ​​the second image data are combined so that the area of ​​the low-reliability area is minimized, and for the valid area, strategic importance,A reliability evaluation-based spatial information update system configured to determine an update priority by analyzing one or more of the possibility of change and update history, selectively update the valid area based on the update priority, verify whether spatial objects within the updated valid area maintain logical consistency with spatial objects of the existing vector map after the update of the valid area is completed, store the result of the verification of logical consistency in conjunction with the integrated reliability score and the determination result of the valid area, and retrain the first model using the stored data.

Images