Digital space construction and display method based on IR engine

Abstract

The invention discloses a digital space construction and display method based on an IR engine, and the method comprises the following steps: S1, collecting three-dimensional model data of a target scene, thereby constructing a retrieval index by adopting a B + tree data structure, supporting batch data parallel retrieval, enabling single data retrieval response time to adapt to the scene data scale and real-time requirements, and obtaining a retrieval result; therefore, the retrieval accuracy is remarkably improved, and meanwhile, the response speed is stable and efficient; the limitation of simple matching of keywords is broken through, the semantic similarity is calculated through an S = M / T quantization formula, meanwhile, a secondary verification mechanism is designed for polysemy words and ambiguous tags, and the matching accuracy is improved; the problem of positioning errors caused by fixed coordinate mapping is optimized, key nodes such as a scene center point and functional area boundary points are automatically recognized, an actual coordinate system is used as reference calibration, node positioning errors adapt to spatial precision requirements, meanwhile, regional semantic tags are associated, and the stability and data relevance of a spatial frame are considered.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of digital twin technology, and in particular to a digital space construction and display method based on an IR engine. BACKGROUND

[0002] With the rapid development of digital twin, smart city and other fields, digital space construction and display technology has become the core support for connecting the physical and virtual worlds. It realizes monitoring, analysis and interaction by digital replication and dynamic presentation of physical scenes, and has broad application prospects in urban planning, industrial operation and other fields.

[0003] In the process of digital space construction and display, the existing technology mostly uses hash table, ordinary binary tree and other traditional index structures, which are difficult to adapt to multi-dimensional and large-scale data retrieval requirements. Hash table has fast one-dimensional retrieval, but it is easy to have hash collision when multi-dimensional correlation occurs, which reduces the accuracy; ordinary binary tree is difficult to maintain balance, and the retrieval response becomes slow when the data size expands, which cannot meet the real-time requirements. In terms of semantic similarity matching, the existing technology mostly relies on keyword complete matching, and lacks deep semantic mining; some technologies introduce semantic calculation, but due to model simplification, there is no polysemy word and ambiguity label verification mechanism, so the matching accuracy is insufficient, and it is difficult to realize accurate data association.

[0004] In terms of space framework construction, the existing technology lacks intelligent identification of key nodes, uses fixed coordinate mapping and does not combine actual coordinate system calibration, resulting in large positioning error of key nodes such as scene center point and functional area boundary point, affecting the stability and accuracy of the space framework.

[0005] SUMMARY

[0006] The present application aims to at least solve one of the technical problems in the related art to some extent.

[0007] To this end, the present application aims to provide a digital space construction and display method based on an IR engine, which uses B+ tree data structure to construct retrieval index, supports batch data parallel retrieval, and adapts the scene data size and real-time requirements in single data retrieval response time, thereby significantly improving the retrieval accuracy while ensuring stable and efficient response speed; breaking through the limitation of simple keyword matching, calculating semantic similarity through the formula S=M / T, and designing a secondary verification mechanism for polysemy words and ambiguity labels to improve the matching accuracy; optimizing the positioning error caused by fixed coordinate mapping, automatically identifying key nodes such as scene center point and functional area boundary point, and calibrating them based on actual coordinate system, so that the node positioning error adapts to the spatial precision requirements, and the regional semantic labels are associated, taking into account the stability of the space framework and the data association.

[0008] To achieve the above object, the application provides a digital space construction and display method based on an IR engine, comprising the following steps:

[0009] S1: collecting three-dimensional model data, semantic label data and real-time dynamic data of a target scene, performing format standardization processing, invalid interference data filtering and data compression on the collected data;

[0010] S2: importing the preprocessed data into the IR engine, constructing a B+ tree data structure retrieval index, and setting the retrieval rule as keyword retrieval combined with semantic similarity matching, wherein the semantic similarity matching adopts a quantitative calculation method based on keyword matching, and the formula is: S = M / T, wherein S is the semantic similarity (0≤S≤1), M is the number of matched keywords, and T is the total number of keywords; the target data is retrieved in real time by the IR engine and an associated relationship is established;

[0011] S3: based on the associated data output by the IR engine, a space framework is built using a Cartesian coordinate system, a basic scene model is generated through polygon mesh modeling technology, model lightweight is realized by combining multi-level detail technology, a dynamic data driving module is integrated, and real-time association of dynamic data and the basic scene model is realized;

[0012] S4: adapting the constructed digital space model to a display terminal, supporting three interaction modes of touch, voice and handle, responding to interaction instructions in real time through the IR engine, dynamically adjusting the space view angle, model state and data display dimension, and synchronously outputting the space roaming path planning result.

[0013] In addition, according to the above-mentioned application, the following additional technical features can also be possessed:

[0014] Specifically, in S1, the three-dimensional model data collection adopts a combination of laser radar scanning and image three-dimensional reconstruction, the laser radar scanning precision is dynamically adjusted according to the specific precision requirements of the target scene, and the image three-dimensional reconstruction ensures to meet the reconstruction precision standard through optimizing the pixel matching algorithm; the semantic label data is generated by combining manual annotation and machine automatic classification, and manual annotation accounts for 30%-50% of the proportion, which is used to calibrate the machine automatic classification result.

[0015] Specifically, in S2, the B+ tree data structure retrieval index supports parallel retrieval of batch data, the response time of single data retrieval is adapted to the data size of the target scene and the real-time retrieval requirements, and the accuracy rate of the retrieval result meets the actual application requirements.

[0016] Specifically, in the S3, the key nodes of the target scene are automatically identified when the spatial framework is built, including the scene center point, the main functional area boundary point, and the high-frequency interaction point. The node coordinates are based on the actual coordinate system of the scene, and the error is controlled within a range that is compatible with the precision requirements associated with the space. The semantic label data of the corresponding area is also associated.

[0017] Specifically, in the S3, model lightening is achieved through multi-level detail technology. The model is divided into three levels (high, medium, and low). The high-detail level is used for close-up display, and more polygons are retained to ensure display effect. The low-detail level is used for long-range display, and the number of polygons is reduced to reduce rendering load.

[0018] Specifically, in the S3, the dynamic data driving module supports multi-source dynamic data access, including temperature, humidity, and other environmental sensor data and device operating status data. Data is accessed through wired or wireless transmission.

[0019] Specifically, in the S4, dynamic display includes a data visualization module. Real-time dynamic data is superimposed on the corresponding model position in the form of floating numerical values or simple charts. The data update delay is consistent with the original data collection rhythm, meeting the real-time requirements.

[0020] Specifically, in the S2, when the semantic similarity is matched, ambiguous words and ambiguous labels are verified again: when the semantic similarity S obtained by basic matching is greater than or equal to 80%, the matching result is directly confirmed; when 50% ≤ S < 80%, further filtering is performed in combination with the type attribute of the target scene to improve the accuracy of semantic matching.

[0021] The beneficial effects of the digital space construction and display method based on the IR engine are as follows:

[0022] 1. B+ tree data structure is used to build search index, supporting batch data parallel search, and single data search response time is adapted to scene data size and real-time demand, thereby significantly improving search accuracy and ensuring stable and efficient response speed.

[0023] 2. Breakthrough simple keyword matching limitations, calculate semantic similarity through S = M / T formula, and design secondary verification mechanism for ambiguous words and ambiguous labels to improve matching accuracy.

[0024] 3. Optimize positioning error caused by fixed coordinate mapping, automatically identify key nodes such as scene center point and functional area boundary point, calibrate based on actual coordinate system, node positioning error adapts to spatial precision requirements, and associate regional semantic labels, taking into account the stability of the spatial framework and data correlation. BRIEF DESCRIPTION OF DRAWINGS

[0025] The above and / or additional aspects and advantages of the present application will become apparent and more readily appreciated from the following description, taken in conjunction with the following drawings of exemplary embodiments of the present application, wherein:

[0026] Figure 1 A schematic diagram of the data acquisition and preprocessing process of the present application is shown in Figure 1.

[0027] Figure 2 A schematic diagram of the IR engine index construction and data retrieval process of the present application is shown in Figure 2. DETAILED DESCRIPTION

[0028] In order to make the technical means, creative features, purposes and effects achieved by the present application easy to understand, the following further describes the present application in conjunction with specific drawings. It should be noted that the embodiments and features in the embodiments of the present application can be combined with each other without conflict.

[0029] The present application is further described in detail below in conjunction with the drawings.

[0030] As shown in Figures 1 and 2, a digital space construction and display method based on an IR engine according to an embodiment of the present application includes the following steps: Figure 1 and Figure 2 As shown in Figures 1 and 2, a digital space construction and display method based on an IR engine according to an embodiment of the present application includes the following steps:

[0031] S1: Collecting three-dimensional model data, semantic label data and real-time dynamic data of a target scene, performing format standardization processing, invalid interference data filtering and data compression on the collected data;

[0032] It should be noted that format standardization is to eliminate the format differences of different data sources so that the data can be compatible with the IR engine and the modeling module.

[0033] Invalid interference data filtering (such as outliers and redundant data) is to reduce data noise and avoid affecting model accuracy and running efficiency.

[0034] Data compression reduces data volume through algorithms, reduces storage and transmission costs, and ensures real-time processing.

[0035] S2: Importing the preprocessed data into the IR engine, constructing a B+ tree data structure retrieval index, and setting the retrieval rule as keyword retrieval combined with semantic similarity matching, wherein the semantic similarity matching adopts a quantitative calculation method based on keyword matching, and the formula is: S = M / T, where S is the semantic similarity (0≤S≤1), M is the number of matched keywords, and T is the total number of keywords. The IR engine retrieves the target data in real time and establishes an association relationship;

[0036] It should be noted that the core function of the IR engine is to quickly retrieve and associate data. The B+ tree is selected as the index structure based on its data organization characteristics: the B+ tree is a balanced multi-way search tree, all data is stored in the leaf nodes, and the leaf nodes are linked in order, which supports both fast positioning of single data and range query and parallel retrieval of batch data, significantly improving the accuracy of retrieval while ensuring stable and efficient response speed.

[0037] The rule of keyword retrieval + semantic similarity matching ensures both the accuracy of retrieval (keyword) and the flexibility (semantic association). The semantic similarity formula S = M / T quantifies the proportion of matching keywords, converts fuzzy semantic association into a calculable value, and makes the matching result more controllable (the closer S is to 1, the higher the matching degree).

[0038] Real-time retrieval and association is to form a logical link between data, providing support for associated data for subsequent modeling.

[0039] S3: Based on the associated data output by the IR engine, a space framework is built using the Cartesian coordinate system, a basic scene model is generated using polygon mesh modeling technology, multi-level detail technology is used to realize model lightweight, a dynamic data driven module is integrated to realize real-time association of dynamic data and basic scene model;

[0040] It should be noted that the space framework is built using the Cartesian coordinate system to provide a unified coordinate reference for the digital space and ensure the accuracy of model position and scale. Polygon mesh modeling technology quickly constructs the geometric form of complex scenes by splicing simple polygons such as triangles and quadrilaterals, and is the mainstream and efficient solution for three-dimensional modeling.

[0041] Multi-level detail technology (LOD) sets different precision levels for the model (low precision for distant objects and high precision for close objects), reduces hardware computing power consumption without affecting visual effects, and realizes lightweight.

[0042] The dynamic data driven module reads dynamic data in real time and updates the model state (such as object motion and data parameter changes) in real time, keeping the digital space synchronized with the real scene.

[0043] S4: The built digital space model is adapted to the display terminal, supporting touch, voice and handle interaction, responding to interactive commands in real time through the IR engine, dynamically adjusting the space perspective, model state and data display dimension, and synchronously outputting the space roaming path planning result.

[0044] It should be noted that terminal adaptation is achieved through compatibility optimization to enable the digital space model to run normally on different devices (such as mobile phones, computers and VR devices), expanding the application scenarios.

[0045] Touch, voice, and handle three interaction modes cover different use scenarios (touch is suitable for mobile devices, and handle is suitable for VR / game scenarios), and improve user operation convenience.

[0046] The IR engine responds to the interaction instruction in real time because the engine has established a data correlation index, can quickly locate the corresponding model or data of the instruction, and realizes low delay of instruction-response.

[0047] In an embodiment of the present application, the three-dimensional model data collection in S1 adopts a combination of laser radar scanning and image three-dimensional reconstruction, the scanning precision of the laser radar is dynamically adjusted according to the specific precision requirements of the target scene, and the image three-dimensional reconstruction ensures to meet the reconstruction precision standard through optimizing the pixel matching algorithm; the semantic label data is generated by combining manual annotation and machine automatic classification, and the manual annotation sets a proportion of 30%-50% for calibrating the machine automatic classification result.

[0048] It should be noted that the three-dimensional model data collection adopts laser radar combined with image three-dimensional reconstruction, which essentially utilizes the complementarity of the two technologies: the laser radar obtains the three-dimensional coordinates of the scene by emitting laser pulses, has the advantages of high precision (millimeter level) and anti-light interference, but has weak ability to capture detailed textures; the image three-dimensional reconstruction based on multi-view image pixel feature matching can supplement the texture details, but the precision is easily affected by light and occlusion. The combination of the two can realize double protection of geometric precision and texture details.

[0049] The precision requirements of different scenes are significantly different (such as millimeter level for industrial equipment and centimeter level for outdoor squares), and the dynamic adjustment of the scanning precision of the laser radar can avoid the waste of cost caused by excessive collection (such as equipment loss and data redundancy).

[0050] The semantic label generation adopts a collaborative mode of machine automatic classification and manual calibration, and the underlying logic is to balance efficiency and accuracy: machine classification can realize fast generation of large-scale labels, but is prone to errors for ambiguous scenes (such as similar objects and complex semantics); manual annotation accounts for 30%-50%, and is used to calibrate samples with low confidence of machine classification. Machine generates efficiently, and human corrects accurately.

[0051] In an embodiment of the present application, the B+ tree data structure retrieval index in S2 supports parallel retrieval of batch data, the retrieval response time of a single data is adapted to the data size and real-time retrieval requirements of the target scene, and the accuracy of the retrieval result meets the actual application requirements.

[0052] It should be noted that the retrieval response time of a single data is adapted to the data size, which is actually dynamic resource allocation: when the data size increases (such as from 100,000 to 10 million), the IR engine will automatically expand the index node and optimize the memory cache strategy to ensure that the response time does not increase linearly with the size.

[0053] In an embodiment of the present application, in S3, when the space framework is built, the key nodes of the target scene are automatically identified, including the scene center point, the main function area boundary point, and the high-frequency interaction point. The node coordinates are based on the actual coordinate system of the scene, and the error is controlled within a range that is adapted to the precision requirement associated with the space, and the semantic label data of the corresponding area is associated.

[0054] It should be noted that the core of the space framework building is to establish the coordinate mapping of the physical space-digital space, and the key nodes (center point, function area boundary point, and high-frequency interaction point) are the anchor points of the mapping: the center point is the origin of the space coordinate system, the function area boundary point defines the space partition, and the high-frequency interaction point (such as the device operation panel and the navigation entrance) is the core object of user interaction.

[0055] The node coordinates are based on the actual coordinate system of the scene, and the essence is to ensure the proportional mapping of the digital space and the physical space, and the error is controlled to adapt to the precision requirement (such as industrial scene error ≤3mm, commercial scene ≤10mm), so as to avoid the misplacement of interaction caused by coordinate offset (such as the mismatch between the digital model position and the physical device position).

[0056] In an embodiment of the present application, in S3, model lightening is realized through multi-level detail technology, and the model is divided into three levels (high, medium, and low). The high-detail level is used for close-range display, and more polygons are retained to ensure the display effect; the low-detail level is used for long-range display, and the number of polygons is reduced to reduce the rendering load.

[0057] Specifically, the core of model lightening is to allocate computing resources on demand: the rendering load of a three-dimensional model is positively correlated with the number of polygons (for every increase of 100,000 polygons, the rendering time increases by about 20ms), and the user's perception of model details changes with distance (close-range needs details, and long-range only needs outlines).

[0058] The division of the three-level detail level is based on the visual perception threshold: the high-detail level (polygons ≥100,000) retains details such as textures and chamfers, and meets the close-range observation (such as <5m); the medium-detail level (30-100,000) simplifies secondary details (such as removing small protrusions), and adapts to medium-distance (5-20m); the low-detail level (≤30,000) only retains the core outline, and adapts to long-distance (>20m), and the level is switched in real time through multi-level detail technology.

[0059] In an embodiment of the present application, in S3, the dynamic data driving module supports multi-source dynamic data access, and the multi-source dynamic data includes environmental sensor data such as temperature and humidity and device running state data; the data is accessed through wired or wireless transmission.

[0060] It should be noted that various types of real-time changing states in the physical scene (such as environmental temperature and humidity, equipment operating parameters) are converted into electrical signals or digital signals by data acquisition devices (sensors, controllers), and after the module receives these signals, it drives the digital space model to make corresponding feedback (such as local color change of the model when the temperature exceeds the standard, model status icon update when the equipment stops) according to the preset rules (such as the binding relationship between data and model elements, state threshold triggering conditions).

[0061] When wired mode is adopted, the advantages are high stability, low delay, and large bandwidth, which are suitable for transmitting high-frequency and critical data (such as real-time equipment operating parameters, which require millisecond-level response). For example, the running state data of the machine tool in the industrial workshop can avoid wireless signal interference through wired connection, meeting the real-time monitoring requirements.

[0062] When wireless mode is adopted, the advantages are flexible deployment and no need for wiring, which are suitable for transmitting low-frequency and non-critical data (such as environmental temperature and humidity, which can be updated every minute) or mobile device data (such as the position information of the inspection robot). For example, the environmental sensor data of the outdoor square can reduce the wiring cost through wireless transmission, and the low-power consumption characteristics are suitable for battery-powered devices.

[0063] In an embodiment of the present application, the dynamic display in S4 includes a data visualization module, and real-time dynamic data is superimposed on the corresponding model position in the form of a floating numerical value or a simple chart. The data update delay is consistent with the original data acquisition rhythm, meeting the real-time requirements.

[0064] It should be noted that dynamic data (such as equipment temperature, human flow density) needs to be linked with the digital space model, and the core is the real-time mapping of space-data. The data visualization module associates dynamic data with model positions to ensure that the data display position is consistent with the physical object.

[0065] The logic of matching the data update delay with the acquisition rhythm is the real-time balance: if the original data is collected every 1 second, the update delay is controlled to be ≤500ms (i.e. the display is completed before the next collection), avoiding the lag of data display behind the actual state (such as the temperature has exceeded the standard but the digital space has not been updated in time).

[0066] In an embodiment of the present application, when the semantic similarity is matched in S2, the polysemous words and ambiguous labels are subjected to secondary verification: when the semantic similarity S obtained by the basic matching is ≥80%, the matching result is directly confirmed; when 50%≤S<80%, further screening is performed in combination with the type attribute of the target scene, so as to improve the accuracy of semantic matching.

[0067] It should be noted that the core challenge of semantic similarity matching is polysemous words / ambiguous labels (such as interface, which refers to equipment connection port in industrial scene, and software interface in IT scene), and the secondary verification is based on the double filtering of threshold + scene attribute.

[0068] When S≥80%, the matching keywords are highly coincident (e.g., high temperature alarm and temperature exceeding alarm), and the probability of false matching is less than 5%, which can be directly confirmed;

[0069] When 50%≤S<80%, the keywords are partially coincident (e.g., equipment failure and mechanical failure), and need to be screened in combination with scene attributes (e.g., industrial workshop preferentially matches mechanical failure) to reduce ambiguity interference.

[0070] It should be noted that in this document, the terms "comprising", "containing", or any other variant thereof are intended to cover non-exclusive inclusions, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or inherent to such a process, method, article or device.

[0071] The above describes the present application and its embodiments, which is not restrictive, and the drawings only show one of the embodiments of the present application, and the actual structure is not limited thereto. In summary, if a person skilled in the art is inspired by it, without departing from the purpose of the present application, without creative design, similar structure and embodiments of the technical solution can be designed, which should belong to the protection scope of the present application.

Claims

1. A method for constructing and displaying digital space based on an IR engine, characterized in that, Includes the following steps: S1: Collect 3D model data, semantic tag data and real-time dynamic data of the target scene, and perform format standardization, invalid interference data filtering and data compression on the collected data; S2: Import the preprocessed data into the IR engine, build a B+ tree data structure retrieval index, and set the retrieval rules to keyword retrieval combined with semantic similarity matching. The semantic similarity matching adopts a quantitative calculation method based on keyword matching, with the formula: S = M / T, where S is the semantic similarity (0≤S≤1), M is the number of matched keywords, and T is the total number of keywords. The IR engine retrieves the target data in real time and establishes the association relationship. S3: Based on the associated data output by the IR engine, a spatial framework is built using the Cartesian coordinate system. A basic scene model is generated through polygon mesh modeling technology. Multi-level detail technology is combined to achieve model lightweighting. A dynamic data-driven module is integrated to realize the real-time association between dynamic data and the basic scene model. S4: Adapts the constructed digital space model to the display terminal, supports three interaction methods: touch, voice and gamepad, responds to interaction commands in real time through the IR engine, dynamically adjusts the spatial perspective, model status and data display dimensions, and outputs the spatial roaming path planning results simultaneously.

2. The method according to claim 1, characterized in that, In S1, the 3D model data acquisition adopts a combination of LiDAR scanning and image 3D reconstruction. The LiDAR scanning accuracy is dynamically adjusted according to the specific accuracy requirements of the target scene, while the image 3D reconstruction ensures that the reconstruction accuracy standard is met by optimizing the pixel matching algorithm. The semantic label data is generated by a combination of manual annotation and automatic machine classification. The manual annotation is set to a proportion of 30%-50% to calibrate the automatic machine classification results.

3. The method according to claim 1, characterized in that, The B+ tree data structure retrieval index in S2 supports parallel retrieval of batch data. The response time for retrieving a single piece of data is adapted to the data scale and real-time retrieval requirements of the target scenario, and the accuracy of the retrieval results meets the needs of practical applications.

4. The method according to claim 1, characterized in that, When constructing the spatial framework in S3, key nodes of the target scene are automatically identified, including the scene center point, the boundary points of the main functional areas, and high-frequency interaction points. The node coordinates are based on the actual coordinate system of the scene, and the error is controlled within a range that is compatible with the accuracy requirements of spatial association. Semantic tag data of the corresponding area is also associated.

5. The method according to claim 1, characterized in that, The model lightweighting in S3 is achieved through multi-level detail technology, which divides the model into three levels (high, medium, and low). The high level of detail is used for close-up display, retaining more polygons to ensure the display effect; the low level of detail is used for distant view display, reducing the number of polygons to reduce the rendering load.

6. The method according to claim 1, characterized in that, The dynamic data driving module in S3 supports multi-source dynamic data access, including environmental sensor data such as temperature and humidity, as well as equipment operating status data; the data is accessed via wired or wireless transmission.

7. The method according to claim 1, characterized in that, The dynamic display in S4 includes a data visualization module, where real-time dynamic data is superimposed on the corresponding model position in the form of floating values ​​or simple charts. The data update delay is consistent with the original data collection rhythm to meet the real-time requirements.

8. The method according to claim 1, characterized in that, In the semantic similarity matching process in S2, a secondary verification is performed on polysemous words and ambiguous tags: when the semantic similarity S obtained from the basic matching is ≥80%, the matching result is directly confirmed; when 50%≤S<80%, further filtering is performed in combination with the type attributes of the target scene to improve the accuracy of semantic matching.

Images