AI recognition-based space accurate mapping method and system

Abstract

This invention discloses a spatial precision mapping method and system based on AI recognition, relating to the field of spatial mapping. The method includes: using a preset judgment method to determine whether a first visual environment information set meets preset conditions, obtaining first mapping information for a first location; moving the mapping device to obtain a second visual environment information set; determining whether the second visual environment information set meets preset conditions, generating second mapping information; inputting the first location, the first mapping information, and the second mapping information into a mapping device location analysis model to obtain location information for the second location; continuing to move the mapping device to obtain multiple mapping information sets and location information for multiple locations; merging the multiple mapping information sets based on the location information for multiple locations to obtain total mapping information for the target indoor environment. This solves the technical problem of low accuracy in spatial mapping in existing technologies, leading to low quality spatial mapping.

Description

Technical Field

[0001] This invention relates to the field of spatial surveying and mapping, specifically to a method and system for precise spatial surveying and mapping based on AI recognition. Background Technology

[0002] Space mapping plays a vital role in location services, national defense, aerospace, and other fields. By measuring and analyzing the components and relationships of space, and exploring its evolutionary patterns, it helps to comprehensively understand the spatial situation, enabling full-scale resource measurement, status awareness, and security management. As space becomes increasingly complex, the difficulty of space mapping continues to rise, and how to conduct effective and accurate space mapping has attracted widespread attention.

[0003] Existing technologies suffer from low accuracy in indoor space mapping, resulting in poor mapping quality. Summary of the Invention

[0004] This application provides a method and system for precise spatial mapping based on AI recognition. It solves the technical problem of low accuracy in existing spatial mapping technologies, which leads to low quality of spatial mapping.

[0005] In view of the above problems, this application provides a method and system for accurate spatial mapping based on AI recognition.

[0006] In a first aspect, this application provides a spatial precision mapping method based on AI recognition, wherein the method includes: at a first location in a target indoor environment, the mapping device acquires a first set of visual environment information through a visual camera device; using a preset judgment method, it is determined whether the first set of visual environment information meets preset conditions; if so, first mapping information for the first location is generated based on the first set of visual environment information; if not, a first set of radar environment information for the first location is acquired through a lidar device, and first mapping information for the first location is generated; the mapping device is moved, and a second set of visual environment information is acquired through the visual camera device. Using the preset judgment method, determine whether the second visual environment information set meets the preset conditions. If yes, generate second mapping information based on the second visual environment information set. If no, obtain the second radar environment information set collected by the lidar device and generate second mapping information. Construct a mapping device position analysis model. Input the first position, first mapping information, and second mapping information into the mapping device position analysis model to obtain the position information of the second position. Continue to move the mapping device to obtain multiple mapping information and position information of multiple positions. Based on the position information of the multiple positions, merge the multiple mapping information to obtain the total mapping information of the target indoor environment.

[0007] Secondly, this application also provides a spatial precision mapping system based on AI recognition, wherein the system includes: a first visual environment information acquisition module, which is used to acquire a first set of visual environment information at a first location in a target indoor environment by means of a mapping device through a visual camera device; a first judgment module, which is used to determine whether the first set of visual environment information meets preset conditions using a preset judgment method; if yes, then generating first mapping information for the first location based on the first set of visual environment information; if no, then acquiring a first set of radar environment information for the first location by means of a lidar device and generating first mapping information for the first location; a second visual environment information acquisition module, which is used to move the mapping device and acquire a second set of visual environment information through the visual camera device; and a second judgment module. The second judgment module is used to determine whether the second visual environment information set meets the preset conditions using the preset judgment method. If yes, it generates second mapping information based on the second visual environment information set; if no, it obtains the second radar environment information set based on the laser radar device and generates second mapping information. The construction module is used to construct a mapping device position analysis model. The second position information acquisition module is used to input the first position, the first mapping information, and the second mapping information into the mapping device position analysis model to obtain the position information of the second position. The mapping module is used to continue moving the mapping device to obtain multiple mapping information and position information of multiple positions. The total mapping information acquisition module is used to merge the multiple mapping information based on the position information of the multiple positions to obtain the total mapping information of the target indoor environment.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0009] The system collects environmental information about the target's indoor environment from a first location to obtain a first visual environmental information set. Using a preset judgment method, it determines whether the first visual environmental information set meets preset conditions. If yes, it generates first mapping information for the first location based on the first visual environmental information set; otherwise, it obtains a first radar environmental information set for the first location based on data collected by a lidar device and generates first mapping information for the first location. The system then moves the mapping device to collect a second visual environmental information set using a visual camera device. Again, using a preset judgment method, it determines whether the second visual environmental information set meets preset conditions. If yes, it generates second mapping information based on the second visual environmental information set; otherwise, it obtains a second radar environmental information set based on data collected by a lidar device and generates second mapping information. A mapping device position analysis model is constructed. The first location, first mapping information, and second mapping information are input into the mapping device position analysis model to obtain the position information for the second location. The system continues to move the mapping device to obtain multiple mapping information sets and position information for multiple locations. Based on the position information for multiple locations, the multiple mapping information sets are merged to obtain the total mapping information for the target's indoor environment. It has achieved the technical effect of realizing intelligent and precise spatial mapping, improving the accuracy and quality of spatial mapping, and laying the foundation for the further precision development of spatial mapping. Attached Figure Description

[0010] Figure 1 This is a flowchart illustrating a spatial precision mapping method based on AI recognition, as described in this application.

[0011] Figure 2 This is a schematic diagram of the process for generating first mapping information of a first location in a spatial precision mapping method based on AI recognition according to this application.

[0012] Figure 3 This is a schematic diagram of the structure of a spatial precision mapping system based on AI recognition, as described in this application.

[0013] Explanation of reference numerals in the attached drawings: First visual environment information acquisition module 11, first judgment module 12, second visual environment information acquisition module 13, second judgment module 14, construction module 15, second location information acquisition module 16, surveying module 17, and total surveying information acquisition module 18. Detailed Implementation

[0014] This application provides a spatial precision mapping method and system based on AI recognition. It solves the technical problem of low accuracy in existing spatial mapping technologies, which leads to low mapping quality. It achieves the technical effect of realizing intelligent and precise spatial mapping, improving the accuracy and quality of spatial mapping, and laying the foundation for further precision development in spatial mapping.

[0015] Example 1

[0016] Please see the appendix Figure 1 This application provides a spatial precision mapping method based on AI recognition, wherein the method is applied to a mapping device based on AI recognition for spatial precision mapping, the mapping device including a visual camera device and a lidar device, and the method specifically includes the following steps:

[0017] Step S100: At the first location in the target indoor environment, the mapping device acquires a first set of visual environment information through the visual camera device;

[0018] Specifically, a surveying device is placed at a first location within the target indoor environment. The visual camera within the surveying device acquires images of the target indoor environment, obtaining a first set of visual environment information. The target indoor environment can be any indoor environment used for intelligent spatial surveying with the surveying device. The first location can be any location within the target indoor environment. The surveying device includes a visual camera and a lidar device. The visual camera can be a machine vision camera such as a CCD camera or a CMOS camera, as used in existing technologies. The first set of visual environment information includes image data information corresponding to the target indoor environment obtained by placing the surveying device at the first location within the target indoor environment. This achieves the technical effect of placing the surveying device at the first location within the target indoor environment, acquiring images of the target indoor environment using the visual camera, and obtaining a first set of visual environment information, laying the foundation for subsequently generating first surveying information for the first location.

[0019] Step S200: Using a preset judgment method, determine whether the first visual environment information set meets the preset conditions. If yes, generate the first mapping information of the first location based on the first visual environment information set. If no, obtain the first radar environment information set of the first location based on the laser radar device and generate the first mapping information of the first location.

[0020] Further details are attached. Figure 2 As shown, step S200 of this application further includes:

[0021] Step S210: Based on the first visual environment information set, obtain the first visual distance between the mapping device and the preset position of the target indoor environment;

[0022] Step S220: Obtain a preset distance threshold and use the preset distance threshold as the preset condition;

[0023] Step S230: Determine whether the first visual distance is less than the preset distance threshold. If yes, generate the first mapping information of the first location based on the first visual environment information set. If no, collect the first radar environment information set of the first location through the lidar device and generate the first mapping information of the first location.

[0024] Specifically, based on a first visual environment information set, the distance information between the mapping device and a preset position in the target indoor environment is determined to obtain the first visual distance. Further, it is determined whether the first visual distance is less than a preset distance threshold. If the first visual distance is less than the preset distance threshold, the accuracy of the first visual environment information set is relatively high, and the first mapping information for the first position is generated based on the first visual environment information set. If the first visual distance is not less than the preset distance threshold, the accuracy of the first visual environment information set is relatively low, and the lidar device in the mapping device is activated to collect radar information from the target indoor environment, obtaining a first radar environment information set, and generating the first mapping information for the first position based on the first radar environment information set. Since lidar consumes a lot of energy during mapping, a visual camera device is preferred for collecting environmental information. When the visual environment information is inaccurate, the lidar device is used to collect and map the environmental information, reducing mapping costs and ensuring mapping accuracy.

[0025] The first visual distance includes distance information between the mapping device and a preset position in the target indoor environment under a first visual environment information set. The preset position in the target indoor environment is preferably the ceiling position. The preset distance threshold includes distance threshold information between the mapping device and the preset position in the target indoor environment, determined by adaptive settings. The preset conditions include the preset distance threshold. The lidar device can be a single-line lidar, multi-line lidar, 3D lidar, etc., as per existing technologies. The first lidar environment information set includes data information such as the spatial structure and geometric shape of spatial objects in the target indoor environment obtained when the lidar device is located at a first position in the target indoor environment. This achieves the technical effect of improving the accuracy of spatial mapping by judging whether the first visual environment information set meets the preset conditions to obtain accurate and reliable first mapping information for the first position.

[0026] Step S300: Move the surveying device and acquire a second set of visual environment information through the visual camera device;

[0027] Step S400: Using the preset judgment method, determine whether the second visual environment information set meets the preset conditions. If yes, generate second mapping information based on the second visual environment information set. If no, obtain the second radar environment information set based on the laser radar device and generate second mapping information.

[0028] Specifically, based on a first location within the target indoor environment, the mapping device is randomly moved, and then images of the target indoor environment are acquired using a visual camera to obtain a second visual environment information set. Further, it is determined whether the second visual environment information set meets preset conditions; that is, based on the second visual environment information set, the distance information between the mapping device and a preset location within the target indoor environment is determined to obtain a second visual distance. It is then determined whether the second visual distance is less than a preset distance threshold. If the second visual distance is less than the preset distance threshold, the second visual environment information set is set as the second mapping information. If the second visual distance is not less than the preset distance threshold, the lidar device in the mapping device is activated, and radar information is acquired from the target indoor environment using the lidar device to obtain a second radar environment information set, which is then set as the second mapping information. The second visual environment information set includes image data information corresponding to the target indoor environment obtained after the mapping device is randomly moved. The second visual distance includes the distance information between the mapping device and the preset location within the target indoor environment under the second visual environment information set. The second radar environmental information set includes data such as the spatial structure and geometric shape of spatial objects in the target indoor environment obtained after the mapping device is randomly moved. This achieves the technical effect of moving the mapping device, obtaining a second visual environmental information set, determining whether the second visual environmental information set meets preset conditions, obtaining second mapping information, and improving the accuracy of spatial mapping.

[0029] Step S500: Construct a location analysis model for the surveying device;

[0030] Furthermore, step S500 of this application also includes:

[0031] Step S510: Based on the target indoor environment, obtain multiple similar indoor environments;

[0032] Step S520: Based on the surveying device, surveying is performed at multiple locations within the multiple family-like indoor environments to obtain multiple initial position information of samples, multiple initial position surveying information of samples, multiple sample movement position surveying information, and multiple sample movement position information;

[0033] Specifically, based on the target indoor environment, multiple similar indoor environments are identified. Further, based on these multiple similar indoor environments, multiple initial position information for samples is obtained. The mapping device is then placed at each of these initial position information locations to collect indoor environmental information, resulting in multiple initial position mapping information for samples. Simultaneously, based on these initial position information, the mapping device is randomly moved to obtain multiple moving position information for samples. The mapping device is then placed at each of these moving position information locations to collect indoor environmental information, resulting in multiple moving position mapping information for samples. The multiple similar indoor environments include multiple indoor environments similar to the target indoor environment. The multiple initial position information for samples includes multiple arbitrary positions within each of these similar indoor environments. The multiple initial position mapping information for samples includes multiple visual mapping information and multiple radar mapping information corresponding to the multiple similar indoor environments under these initial position information. The multiple moving position information for samples includes multiple moving position information for the mapping device obtained after randomly moving the mapping device based on these initial position information. The multiple sample movement location mapping information includes multiple visual mapping information and multiple radar mapping information corresponding to multiple similar indoor environments, under multiple sample movement location information. This achieves the technical effect of obtaining multiple sample initial position information, multiple sample initial position mapping information, multiple sample movement location mapping information, and multiple sample movement location information by mapping multiple locations within multiple similar indoor environments, providing data support for subsequent construction of a mapping device location analysis model.

[0034] Step S530: Using the initial position information of the multiple samples, the initial position mapping information of the multiple samples, the moving position mapping information of the multiple samples, and the moving position information of the multiple samples, construct the position analysis model of the mapping device.

[0035] Furthermore, step S530 of this application also includes:

[0036] Step S531: Based on the feedforward neural network, construct the position analysis model of the surveying device. The input data of the position analysis model of the surveying device are the position information of the initial position, the surveying information of the initial position and the surveying information of the moving position, and the output data is the position information of the moving position.

[0037] Step S532: Data identification is performed on the initial position information of multiple samples, the initial position mapping information of multiple samples, the moving position mapping information of multiple samples, and the moving position information of multiple samples to obtain the constructed dataset;

[0038] Step S533: Use the constructed dataset to perform iterative supervised training on the location analysis model of the surveying device until the accuracy of the location analysis model of the surveying device meets the preset requirements;

[0039] Step S534: The constructed dataset is used to verify the location analysis model of the surveying device. If the accuracy of the location analysis model of the surveying device meets the preset requirements, the constructed location analysis model of the surveying device is obtained.

[0040] Specifically, multiple initial position information of samples, multiple initial position mapping information of samples, multiple sample movement position mapping information, and multiple sample movement position information are divided and labeled to obtain a constructed dataset. The constructed dataset includes a sample training set and a sample test set. For example, the sample training set includes 70% of the initial position information of multiple samples, 70% of the initial position mapping information of multiple samples, 70% of the movement position mapping information of multiple samples, and 70% of the movement position information of multiple samples randomly. Furthermore, the data information in the sample training set has a corresponding relationship. The sample test set includes multiple initial position information of samples, multiple initial position mapping information of samples, multiple movement position mapping information of multiple samples, and data information other than those in the sample training set from the multiple sample movement position information. Further, the sample training set in the constructed dataset undergoes continuous self-training and iterative supervised training until the accuracy of the mapping device position analysis model meets preset requirements, at which point training ends. Furthermore, the sample test set in the constructed dataset is used as input information to the surveying device location analysis model. The parameters of the surveying device location analysis model are updated and verified. When the accuracy of the surveying device location analysis model meets the preset requirements, the constructed surveying device location analysis model is obtained. The preset requirements include a pre-set accuracy threshold for the surveying device location analysis model. The feedforward neural network is a unidirectional multi-layer structure with an input layer, hidden layer, and output layer, and the neurons in the feedforward neural network are arranged hierarchically. Each neuron is only connected to the neurons in the previous layer, receives the output of the previous layer, and outputs it to the next layer; there is no feedback between layers. The surveying device location analysis model is a feedforward neural network model whose accuracy meets the preset requirements. The input information of the surveying device location analysis model includes the initial position information, the initial position surveying information, and the moving position surveying information; the output information is the moving position information. This achieves the technical effect of obtaining a surveying device location analysis model with high accuracy and strong generalization performance through a feedforward neural network and a constructed dataset, thereby improving the reliability of subsequently obtained location information.

[0041] Step S600: Input the first location, the first mapping information, and the second mapping information into the location analysis model of the mapping device to obtain the location information of the second location;

[0042] Step S700: Continue moving the surveying device to obtain multiple surveying information and location information of multiple locations;

[0043] Specifically, the first location, first mapping information, and second mapping information are used as input information and input into the mapping device location analysis model to obtain the location information of the second location. Further, based on the target indoor environment, the mapping device is moved randomly multiple times, and the target indoor environment is mapped using the mapping device to obtain multiple mapping information. The first location, first mapping information, and multiple mapping information are then used as input information and input into the mapping device location analysis model to obtain location information for multiple locations. The location information of the second location includes the placement location information of the mapping device corresponding to the second mapping information. The multiple mapping information includes multiple visual mapping information and multiple radar mapping information corresponding to the target indoor environment obtained after further random movement of the mapping device. The location information of the multiple locations includes multiple placement location information of the mapping device corresponding to the multiple mapping information. This achieves the technical effect of improving the accuracy and reliability of mapping the target indoor environment by analyzing the first location, first mapping information, and second mapping information through the mapping device location analysis model to obtain the location information of the second location, and by continuing to move the mapping device to obtain multiple mapping information and multiple location information.

[0044] Step S800: Based on the location information of the multiple locations, merge the multiple surveying information to obtain the total surveying information of the target indoor environment.

[0045] Furthermore, step S800 of this application also includes:

[0046] Step S810: Obtain several visual mapping information from the plurality of mapping information;

[0047] Step S820: Based on the plurality of visual mapping information, obtain a plurality of visual distances between the mapping device and the preset position;

[0048] Step S830: Calculate the difference between the plurality of visual distances and the preset distance threshold to obtain a plurality of distance differences;

[0049] Specifically, visual mapping information is extracted based on multiple surveying and mapping data to obtain several visual mapping information sets. Further, based on these visual mapping information sets, the distance information between the surveying device and a preset position in the target indoor environment is determined, obtaining several visual distances. The differences between these visual distances and preset distance thresholds are calculated to obtain several distance difference values. The several visual mapping information sets include multiple image data acquired by the visual camera device from the multiple surveying and mapping data sets. The several visual distances include multiple distances between the surveying device and the preset position in the target indoor environment based on the several visual mapping information sets. The several distance difference values ​​include multiple differences between the several visual distances and the preset distance thresholds. This achieves the technical effect of improving the accuracy and adaptability of subsequent corrections of the several visual mapping information sets by analyzing and calculating several distance difference values ​​from multiple surveying and mapping data sets.

[0050] Step S840: Construct the correction amplitude analysis model;

[0051] Furthermore, step S840 of this application also includes:

[0052] Step S841: Obtain the distance difference between multiple samples of visual mapping information;

[0053] Step S842: Based on the distance differences between the multiple samples, evaluate the accuracy and correction magnitude of the visual mapping information of the multiple samples to obtain the correction magnitude information of the multiple samples;

[0054] Step S843: Randomly select one sample distance difference from the plurality of sample distance differences to construct the first-level classification node of the correction amplitude analysis model;

[0055] Step S844: Randomly select another sample distance difference from the plurality of sample distance differences to construct the secondary classification node of the correction amplitude analysis model;

[0056] Step S845: Continue to construct the multi-level classification nodes of the correction amplitude analysis model, and obtain multiple classification results obtained by the multi-level classification nodes;

[0057] Step S846: Using the multiple sample correction magnitude information, label the multiple classification results to obtain the constructed correction magnitude analysis model.

[0058] Step S850: Input the plurality of distance differences into the correction amplitude analysis model to obtain a plurality of correction amplitude information;

[0059] Step S860: Determine whether the plurality of visual mapping information needs to be corrected. If so, use the plurality of correction amplitude information to correct the plurality of visual mapping information.

[0060] Specifically, visual mapping information is extracted from the initial position mapping information and the moving position mapping information of multiple samples, resulting in multiple sample visual mapping information. This visual mapping information is then analyzed and calculated to obtain multiple sample distance differences. Based on these distance differences, the accuracy and correction magnitude of the sample visual mapping information are evaluated, yielding multiple sample correction magnitude information. Further, multiple sample distance differences are randomly selected to obtain first-level classification nodes. Second-level classification nodes are then randomly selected again. This process is repeated multiple times to obtain multi-level classification nodes. Classification is performed according to the range of the correction magnitude information corresponding to the multi-level classification nodes, resulting in multiple classification results. These results are then labeled according to the correction magnitude information, resulting in a completed correction magnitude analysis model. Finally, several distance differences are used as input information and fed into the correction magnitude analysis model to obtain several correction magnitude information. Next, it is determined whether several visual mapping information needs to be corrected. That is, several radar mapping information from multiple mapping information are compared with the corresponding several visual mapping information. When the deviation between several radar mapping information and the corresponding several visual mapping information is large, several correction amplitude information is used to correct the several visual mapping information.

[0061] The multiple sample visual mapping information includes multiple initial position mapping information and multiple sample movement position mapping information, as well as multiple image data information acquired by a visual camera device. The multiple sample distance differences include multiple differences between multiple visual distances and preset distance thresholds in the multiple sample visual mapping information. The multiple sample correction amplitude information includes the magnitude, direction, and orientation of the correction applied to the multiple sample visual mapping information under the multiple sample distance differences. The first-level classification node includes any sample distance difference among the multiple sample distance differences. The second-level classification node includes any sample distance difference that differs from the first-level classification node among the multiple sample distance differences. The multi-level classification node includes multiple sample distance differences. The multiple classification results include the range of multiple sample correction amplitude information corresponding to the multi-level classification nodes. The correction amplitude analysis model includes multiple classification results and multiple sample correction amplitude information corresponding to the multiple classification results. The several correction amplitude information includes the magnitude, direction, and orientation of the correction amplitude corresponding to several distance differences. The technology achieves the effect of analyzing the correction amplitude of several distance differences through a correction amplitude analysis model to obtain several correction amplitude information. When several visual mapping information needs to be corrected, the visual mapping information is corrected according to the several correction amplitude information, thereby improving the accuracy of several visual mapping information among multiple mapping information, and thus improving the accuracy of spatial mapping.

[0062] Furthermore, step S800 of this application also includes:

[0063] Step S870: Based on the location information of the multiple locations, construct the location topology structure of the target indoor environment;

[0064] Step S880: Based on the location topology, merge the multiple surveying information to obtain preliminary total surveying information of the target indoor environment and a splicing area of ​​multiple surveying information;

[0065] Step S890: In the multiple mapping information splicing areas, noise filtering is performed based on median filtering to obtain the total mapping information of the target indoor environment.

[0066] Specifically, a location topology is constructed based on the location information of a first location, a second location, and multiple locations. Further, the first surveying information, the second surveying information, and multiple surveying information are merged according to the location topology to obtain preliminary overall surveying information of the target indoor environment and a spliced ​​area of ​​multiple surveying information. Median filtering is then applied to the spliced ​​area to filter noise. Combined with the preliminary overall surveying information, the overall surveying information of the target indoor environment is obtained. The location topology includes the location network structure information between the first location, the second location, and the multiple locations. The location topology can be used to characterize the positional and spatial relationships between the first location, the second location, and the multiple locations. The preliminary overall surveying information includes the overall surveying information of the target indoor environment obtained by merging the first surveying information, the second surveying information, and multiple surveying information according to the location topology. The spliced ​​area of ​​multiple surveying information includes the surveying information of the corresponding spliced ​​positions when merging the first surveying information, the second surveying information, and multiple surveying information according to the location topology. Median filtering is a nonlinear signal processing technique that effectively suppresses noise. The basic principle of median filtering is to replace the value of a point in a stitched area of ​​multiple surveying information with the median value of all points in its neighborhood, making the surrounding pixel values ​​closer to the true value. This eliminates isolated noise points in the stitched area, preventing blurring and improving its clarity. The total surveying information of the target indoor environment includes preliminary total surveying information and stitched areas of multiple surveying information after median filtering. This achieves the technical effect of merging first, second, and multiple surveying information according to the location topology to obtain preliminary total surveying information and stitched areas of multiple surveying information for the target indoor environment, and then applying median filtering to these stitched areas to obtain accurate and clear total surveying information, thus improving the quality of spatial surveying.

[0067] In summary, the spatial accurate mapping method based on AI recognition provided in this application has the following technical effects:

[0068] 1. The system acquires images of the target indoor environment from a first location to obtain a first visual environment information set. Using a preset judgment method, it determines whether the first visual environment information set meets preset conditions. If yes, it generates first mapping information for the first location based on the first visual environment information set. If no, it acquires a first radar environment information set for the first location using a lidar device and generates first mapping information for the first location. The system moves the mapping device to acquire a second visual environment information set using a visual camera device. Using a preset judgment method, it determines whether the second visual environment information set meets preset conditions. If yes, it generates second mapping information based on the second visual environment information set. If no, it acquires a second radar environment information set using a lidar device and generates second mapping information. A mapping device position analysis model is constructed. The first location, first mapping information, and second mapping information are input into the mapping device position analysis model to obtain the position information for the second location. The system continues to move the mapping device to obtain multiple mapping information sets and position information for multiple locations. Based on the position information for multiple locations, the multiple mapping information sets are merged to obtain the total mapping information for the target indoor environment. It has achieved the technical effect of realizing intelligent and precise spatial mapping, improving the accuracy and quality of spatial mapping, and laying the foundation for the further precision development of spatial mapping.

[0069] 2. By analyzing the first location, first surveying information, and second surveying information through the surveying device location analysis model, the location information of the second location is obtained. The surveying device is then moved to obtain multiple surveying information and location information of multiple locations, thereby improving the accuracy and reliability of surveying the target indoor environment.

[0070] 3. By using a correction amplitude analysis model to analyze the correction amplitude of several distance differences, several correction amplitude information is obtained. When several visual mapping information needs to be corrected, the visual mapping information is corrected according to the several correction amplitude information, thereby improving the accuracy of several visual mapping information among multiple mapping information, and thus improving the accuracy of spatial mapping.

[0071] Example 2

[0072] Based on the aforementioned AI-based spatial precision mapping method, and using the same inventive concept, this invention also provides an AI-based spatial precision mapping system. The system is applied to a mapping device for AI-based spatial precision mapping, which includes a visual camera and a lidar device. (See attached diagram.) Figure 3 The system includes:

[0073] The first visual environment information acquisition module 11 is used to acquire a first visual environment information set by the surveying device through the visual camera device at a first location in the target indoor environment.

[0074] The first judgment module 12 is used to judge whether the first visual environment information set meets the preset conditions using a preset judgment method. If yes, the first mapping information of the first location is generated based on the first visual environment information set. If no, the first radar environment information set of the first location is obtained by the lidar device and the first mapping information of the first location is generated.

[0075] The second visual environment information acquisition module 13 is used to move the surveying device and acquire a second visual environment information set through the visual camera device.

[0076] The second judgment module 14 is used to determine whether the second visual environment information set meets the preset conditions using the preset judgment method. If yes, the second mapping information is generated based on the second visual environment information set. If no, the second radar environment information set is obtained based on the laser radar device and the second mapping information is generated.

[0077] Construction module 15, which is used to construct a location analysis model of the surveying device;

[0078] The second location information acquisition module 16 is used to input the first location, the first mapping information, and the second mapping information into the location analysis model of the mapping device to obtain the location information of the second location.

[0079] The surveying module 17 is used to continue moving the surveying device to obtain multiple surveying information and multiple location information.

[0080] The total mapping information acquisition module 18 is used to merge the multiple mapping information based on the location information of the multiple locations to obtain the total mapping information of the target indoor environment.

[0081] Furthermore, the system also includes:

[0082] The first visual distance determination module is used to obtain the first visual distance between the surveying device and the preset position of the target indoor environment based on the first visual environment information set.

[0083] A preset condition determination module is used to obtain a preset distance threshold and use the preset distance threshold as the preset condition.

[0084] The first mapping information determination module is used to determine whether the first visual distance is less than the preset distance threshold. If so, it generates the first mapping information of the first location based on the first visual environment information set. If not, it collects the first radar environment information set of the first location through the lidar device and generates the first mapping information of the first location.

[0085] Furthermore, the system also includes:

[0086] A family-based indoor environment determination module is used to obtain multiple family-based indoor environments based on the target indoor environment;

[0087] The sample information determination module is used to perform surveying and mapping at multiple locations within the multiple family-like indoor environments based on the surveying and mapping device, and to obtain multiple initial position information of samples, multiple initial position surveying information of samples, multiple sample movement position surveying information, and multiple sample movement position information.

[0088] The first execution module is used to construct the position analysis model of the mapping device by using the initial position information of the multiple samples, the initial position mapping information of the multiple samples, the moving position mapping information of the multiple samples, and the moving position information of the multiple samples.

[0089] Furthermore, the system also includes:

[0090] The second execution module is used to construct the position analysis model of the surveying device based on the feedforward neural network. The input data of the position analysis model of the surveying device are the position information of the initial position, the surveying information of the initial position and the surveying information of the moving position, and the output data is the position information of the moving position.

[0091] A dataset determination module is used to identify the initial location information of multiple samples, the initial location mapping information of multiple samples, the moving location mapping information of multiple samples, and the moving location information of multiple samples to obtain a constructed dataset.

[0092] An iterative supervised training module is used to perform iterative supervised training on the location analysis model of the surveying device using the constructed dataset until the accuracy of the location analysis model of the surveying device meets the preset requirements.

[0093] The verification module is used to verify the location analysis model of the surveying device using the constructed dataset. If the accuracy of the location analysis model of the surveying device meets the preset requirements, the constructed location analysis model of the surveying device is obtained.

[0094] Furthermore, the system also includes:

[0095] A location topology determination module is used to construct and obtain the location topology of the target indoor environment based on the location information of the multiple locations;

[0096] A mapping information merging module is used to merge multiple mapping information based on the location topology to obtain preliminary total mapping information of the target indoor environment and a splicing area of ​​multiple mapping information.

[0097] A noise filtering module is used to perform noise filtering based on median filtering in the splicing area of ​​multiple surveying information to obtain the total surveying information of the target indoor environment.

[0098] Furthermore, the system also includes:

[0099] A visual mapping information determination module is used to acquire several visual mapping information from the plurality of mapping information;

[0100] A visual distance acquisition module is used to obtain several visual distances between the mapping device and the preset position based on the several visual mapping information.

[0101] A distance difference acquisition module is used to calculate the difference between the plurality of visual distances and the preset distance threshold to obtain a plurality of distance differences;

[0102] The third execution module is used to construct the correction amplitude analysis model;

[0103] A correction amplitude information acquisition module is used to input the plurality of distance differences into the correction amplitude analysis model to obtain a plurality of correction amplitude information.

[0104] A correction processing module is used to determine whether the plurality of visual mapping information needs to be corrected. If so, the plurality of correction amplitude information is used to correct the plurality of visual mapping information.

[0105] Furthermore, the system also includes:

[0106] A sample distance difference determination module is used to obtain multiple sample distance differences from multiple sample visual mapping information.

[0107] A sample correction amplitude information acquisition module is used to evaluate the accuracy and correction amplitude of the visual mapping information of the multiple samples based on the distance differences between the multiple samples, and to obtain multiple sample correction amplitude information.

[0108] A first-level classification node construction module is used to randomly select a sample distance difference from the plurality of sample distance differences to construct the first-level classification node of the correction amplitude analysis model;

[0109] A secondary classification node construction module is used to randomly select a sample distance difference from the plurality of sample distance differences to construct the secondary classification node of the correction amplitude analysis model;

[0110] A module for obtaining multiple classification results is provided, which is used to further construct multi-level classification nodes of the correction magnitude analysis model and obtain multiple classification results obtained by the multi-level classification nodes.

[0111] A labeling module is used to label the multiple classification results using the multiple sample correction magnitude information to obtain the constructed correction magnitude analysis model.

[0112] This application provides a spatial precision mapping method based on AI recognition, wherein the method is applied to a spatial precision mapping system based on AI recognition. The method includes: acquiring images of the target indoor environment from a first location to obtain a first visual environment information set; determining whether the first visual environment information set meets preset conditions using a preset judgment method; if yes, generating first mapping information for the first location based on the first visual environment information set; if no, acquiring a first radar environment information set for the first location from a lidar device and generating first mapping information for the first location; and moving the mapping device to acquire second visual environment information from a visual camera device. The system first sets up a second set of visual environmental information. Using a preset judgment method, it determines whether the second set of visual environmental information meets preset conditions. If yes, it generates second mapping information based on the second set of visual environmental information; otherwise, it obtains a second set of radar environmental information collected by a lidar device and generates second mapping information. A mapping device position analysis model is constructed. The first position, first mapping information, and second mapping information are input into the mapping device position analysis model to obtain the position information of the second position. The mapping device is then moved to obtain multiple sets of mapping information and position information of multiple positions. Based on the position information of multiple positions, the multiple sets of mapping information are merged to obtain the total mapping information of the target indoor environment. This solves the technical problem of low accuracy in spatial mapping in existing technologies, which leads to low quality spatial mapping. It achieves the technical effect of realizing intelligent and precise spatial mapping, improving the accuracy and quality of spatial mapping, and laying the foundation for further precision development in spatial mapping.

[0113] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0114] This specification and accompanying drawings are merely illustrative examples of this application. If any modifications and variations of this invention fall within the scope of this invention and its equivalents, this invention also intends to include such modifications and variations.

Claims

1. A spatial accurate mapping method based on AI recognition, characterized in that, The method is applied to a mapping device for precise spatial mapping based on AI recognition. The mapping device includes a visual camera and a lidar device. The method includes: At a first location in the target indoor environment, the mapping device acquires a first set of visual environment information through the visual camera device; A preset judgment method is used to determine whether the first visual environment information set meets the preset conditions. If yes, the first mapping information of the first location is generated based on the first visual environment information set. If no, the first radar environment information set of the first location is obtained by the lidar device and the first mapping information of the first location is generated. The surveying device is moved, and a second set of visual environment information is acquired through the visual camera device. Using the preset judgment method, it is determined whether the second visual environment information set meets the preset conditions. If yes, second mapping information is generated based on the second visual environment information set. If no, the second radar environment information set is obtained based on the laser radar device, and second mapping information is generated. Construct a location analysis model for surveying equipment; The first location, the first mapping information, and the second mapping information are input into the location analysis model of the mapping device to obtain the location information of the second location. Continue moving the surveying device to obtain multiple surveying information and location information of multiple locations; Based on the location information of the multiple locations, the multiple surveying information are merged to obtain the total surveying information of the target indoor environment; The construction of the mapping device location analysis model includes: Based on the target indoor environment, multiple similar indoor environments are obtained; Based on the surveying device, surveying is performed at multiple locations within the multiple family-like indoor environments to obtain multiple initial position information of samples, multiple initial position surveying information of samples, multiple sample movement position surveying information, and multiple sample movement position information. Using the initial position information of multiple samples, the initial position mapping information of multiple samples, the moving position mapping information of multiple samples, and the moving position information of multiple samples, a position analysis model for the mapping device is constructed, including: Based on a feedforward neural network, a position analysis model for the surveying device is constructed. The input data of the position analysis model for the surveying device includes the position information of the initial position, the surveying information of the initial position, and the surveying information of the moving position. The output data is the position information of the moving position. Data identification is performed on the initial position information of multiple samples, the mapping information of the initial position of multiple samples, the mapping information of the moving position of multiple samples, and the moving position information of multiple samples to obtain the constructed dataset; The constructed dataset is used to perform iterative supervised training on the location analysis model of the surveying device until the accuracy of the location analysis model of the surveying device meets the preset requirements; The constructed dataset is used to verify the location analysis model of the surveying device. If the accuracy of the location analysis model of the surveying device meets the preset requirements, the constructed location analysis model of the surveying device is obtained.

2. The method according to claim 1, characterized in that, A preset judgment method is used to determine whether the first visual environment information set meets preset conditions, including: Based on the first set of visual environment information, the first visual distance between the mapping device and the preset position of the target indoor environment is obtained; Obtain a preset distance threshold and use the preset distance threshold as the preset condition; If the first visual distance is less than the preset distance threshold, then the first mapping information of the first location is generated based on the first visual environment information set; otherwise, the first radar environment information set of the first location is acquired by the lidar device, and the first mapping information of the first location is generated.

3. The method according to claim 1, characterized in that, Based on the location information of the multiple locations, the multiple mapping information are merged, including: Based on the location information of the multiple locations, a location topology structure for obtaining the target indoor environment is constructed; Based on the location topology, the multiple surveying information is merged to obtain preliminary overall surveying information of the target indoor environment and a splicing area of ​​multiple surveying information; In the splicing area of ​​the multiple survey information, noise filtering is performed based on median filtering to obtain the total survey information of the target indoor environment.

4. The method according to claim 2, characterized in that, Before merging the multiple mapping information based on the location information of the multiple locations, the method further includes: Obtain several visual mapping information from the multiple mapping information; Based on the aforementioned visual mapping information, several visual distances between the mapping device and the preset position are obtained; Calculate the difference between the plurality of visual distances and the preset distance threshold to obtain a plurality of distance differences; Construct a correction amplitude analysis model; The plurality of distance differences are input into the correction amplitude analysis model to obtain a plurality of correction amplitude information; Determine whether the plurality of visual mapping information needs to be corrected. If so, use the plurality of correction magnitude information to correct the plurality of visual mapping information.

5. The method according to claim 4, characterized in that, The construction of the correction amplitude analysis model includes: Obtain the distance difference between multiple samples from multiple visual mapping information; Based on the distance differences between the multiple samples, the accuracy and correction magnitude of the visual mapping information of the multiple samples are evaluated to obtain the correction magnitude information of the multiple samples; Randomly select a sample distance difference from the plurality of sample distance differences to construct the first-level classification node of the correction magnitude analysis model; Another sample distance difference is randomly selected from the plurality of sample distance differences to construct the secondary classification node of the correction amplitude analysis model; Continue to construct the multi-level classification nodes of the correction amplitude analysis model, and obtain multiple classification results obtained by the multi-level classification nodes; Using the multiple sample correction magnitude information, the multiple classification results are labeled to obtain the completed correction magnitude analysis model.

6. A spatial precision mapping system based on AI recognition, characterized in that, The system is applied to a mapping device for precise spatial mapping based on AI recognition. The mapping device includes a visual camera and a lidar device. The system includes: The first visual environment information acquisition module is used to acquire a first set of visual environment information by the surveying device through the visual camera device at a first location in the target indoor environment. The first judgment module is used to determine whether the first visual environment information set meets the preset conditions using a preset judgment method. If yes, the first mapping information of the first location is generated based on the first visual environment information set. If no, the first radar environment information set of the first location is obtained by the lidar device and the first mapping information of the first location is generated. The second visual environment information acquisition module is used to move the surveying device and acquire a second visual environment information set through the visual camera device. The second judgment module is used to determine whether the second visual environment information set meets the preset conditions using the preset judgment method. If yes, the second mapping information is generated based on the second visual environment information set. If no, the second radar environment information set is obtained based on the laser radar device and the second mapping information is generated. The construction module is used to construct a location analysis model for the surveying device; The second location information acquisition module is used to input the first location, the first mapping information, and the second mapping information into the location analysis model of the mapping device to obtain the location information of the second location. A surveying module, which is used to continue moving the surveying device to obtain multiple surveying information and multiple location information; The total mapping information acquisition module is used to merge the multiple mapping information based on the location information of the multiple locations to obtain the total mapping information of the target indoor environment.

Images