AI-based artificial intelligence real digital scene hybrid construction method and system
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- 杭州中谦科技有限公司
- Filing Date
- 2024-10-10
- Publication Date
- 2026-07-21
Smart Images
Figure CN119206143B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of scene construction technology, specifically to a method and system for hybrid construction of real-world digital scenes based on AI. Background Technology
[0002] Digital scenario construction is a comprehensive process involving the integration of multiple technologies and fields. It aims to transform and upgrade traditional physical scenarios through digital technology to improve efficiency, reduce costs, and enhance user experience. Application examples of digital scenario construction include: using IoT, big data, and AI technologies to digitally transform and upgrade urban transportation, public safety, and environmental protection, achieving optimized allocation and efficient management of urban resources and improving the quality of life and well-being of urban residents; creating immersive tourism experiences through virtual reality, 3D modeling, and other technologies, allowing tourists to experience the beauty and cultural atmosphere of their destinations from home, increasing tourism appeal and satisfaction; and utilizing IoT and AI technologies to achieve remote monitoring and intelligent diagnosis of medical equipment, improving the efficiency and quality of medical services while reducing medical costs and risks. In the process of hybrid construction of traffic intersections, people often only build the scene information of the intersection's buildings and some obstacles, and then simulate the traffic flow at the intersection. However, they often overlook some details, such as potholes and cracks in the road surface. These potholes and cracks are caused by the constant passage of trucks at traffic intersections, and the road surface cannot withstand the pressure of these trucks for a long time, resulting in cracks and potholes. When cracks and potholes appear on the road surface, they often affect the traffic flow at the intersection. Therefore, we propose a hybrid construction method and system based on AI artificial intelligence and real-world digital scenes. Summary of the Invention
[0003] To address the aforementioned technical issues, this technical solution provides a method and system for hybrid construction of real-world digital scenarios based on AI, thus resolving the problems mentioned above.
[0004] To achieve the above objectives, the technical solution adopted by this invention is: a method for hybrid construction of real-world digital scenes based on AI artificial intelligence, the construction method including: Drones are used to scan and identify traffic intersections to obtain 3D point cloud data; The collected data is preprocessed and uploaded to modeling software to construct the intersection, which is then built into the computer. Obtain the current traffic flow at the intersection, and based on the traffic flow, determine the frequency of container truck appearances; Based on the frequency of the trucks, the stress on the road surface is calculated. Based on the stress, the corresponding road surface potholes and cracks are built on the road surface of the traffic intersection model. Based on the constructed digital intersection scenario, the traffic flow and road surface potholes at the intersection are simulated and predicted, and adjustments are made in the real environment.
[0005] Preferably, the UAV is equipped with a 3D laser scanner. The UAV flies over the intersection and uses the 3D laser scanner to scan the intersection. Based on the position of the intersection point between the laser beam and the object surface, the 3D point cloud data of the intersection is obtained. The obtained data is preprocessed and uploaded to the modeling software to construct the intersection. The 3D point cloud data preprocessing includes point cloud filtering, point cloud key point extraction and point cloud registration.
[0006] Preferably, the intersection construction is based on the Poisson reconstruction algorithm to process the processed data and construct it in the modeling software. An octree is constructed according to the density of the point cloud data, and a function is defined for the octree, with the function form as follows: Where X(q) is an indicator function used to define whether a point q in space is located on the surface of an object M, where q is a point in the three-dimensional point cloud data space, and M is the measured object; A vector field is constructed using point cloud data, the Poisson equation is solved using the vector field, isosurfaces are extracted, and reconstruction is completed. The formula for a vector field is: Where V is the normal vector of all P in the point cloud data. Let p be a point in the point cloud, and w be a point element. Let n be the normal vector of point P, and n be the neighborhood, which is the set of other points in the space surrounding point P. The extracted isosurfaces are integrated with the original point data to form a complete 3D intersection model.
[0007] Preferably, based on historical traffic flow data obtained from intersection monitoring, the traffic flow data of the monitored intersection is obtained through image analysis. Within a unit of time, the number of times a container truck appears at the current intersection is counted, and this data is used as the first set of data. Multiple sets of data samples are obtained as the dataset, and the average frequency of container truck occurrences in the dataset is obtained. The frequency calculation formula is as follows: Where L is the frequency of truck appearance, i is the number of times trucks appear, and t is the unit of time.
[0008] Preferably, the pressure exerted on the road surface by a container truck is calculated. Before the calculation, the weight of the container truck when fully loaded is obtained and estimated based on the container truck's load limit. The calculation formula is as follows: F=G=mo Where G is gravity, m is mass, o is the speed of the truck traveling on the road, o is based on the average speed of the truck when it passes the road, and F is the pressure on the road surface. The formula for calculating the pressure on the road surface is: P is the pressure on the road surface, F is the force on the road surface, and S is the force-bearing area, where the force-bearing area is the area of the road surface that the tire rolls over. The contact surface between the tire and the ground is a rectangle or ellipse, and the force-bearing area is calculated based on the tire width and the predicted contact length.
[0009] Preferably, the occurrence of road surface cracks and potholes is predicted. Factors contributing to pothole formation include the road surface material structure coefficient, the frequency of trucks at intersections, and the construction quality coefficient. The road surface material structure coefficient is obtained based on road construction standards. The pothole assessment formula is as follows: Where R is the pressure coefficient of the road material, L is the frequency of truck occurrence, P is the pressure on the road surface, B is the construction quality coefficient, and H is the probability of potholes appearing on the road surface. The road surface texture effect is built using modeling software, including cracks and potholes.
[0010] Preferably, a deep learning algorithm is used in a computer to construct the effect of cracks and potholes on the traffic road surface based on the obtained road surface pothole probability, and the constructed crack and pothole effect is verified to see if the constructed crack and pothole effect matches the actual scenario.
[0011] Preferably, the features of different datasets are validated, and the validation formula is as follows: Where Z is a given feature Under these conditions, does the dataset input from the road surface belong to the normal crack analogy? This represents the distance threshold between cracks and potholes on the road and the roadside in a real-world scenario. The parameter vector of the model is input with the features of the constructed road surface crack and pothole effect dataset to determine whether the constructed road surface effect matches the actual situation.
[0012] The AI-based digital scene hybrid construction system includes a modeling module, a road surface stress analysis module, a prediction module, and a data acquisition module. The modeling module uses a computer as a platform to build hybrid scenes, the road surface stress analysis module analyzes the stress on the road surface, the prediction module predicts and analyzes the actual intersection conditions based on the constructed digital scene, and the data acquisition module collects data from actual intersections to construct a dataset.
[0013] Preferably, the acquisition module collects real-time traffic intersection conditions based on the UAV and obtains three-dimensional point cloud data of the intersection. The acquisition module also obtains image data of the intersection based on the intersection monitoring and processes the acquired images based on image analysis methods.
[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention uses drones to scan 3D point cloud data of traffic intersections, uploads it to a computer to construct a scene of the current intersection, and obtains the frequency of trucks appearing at the current intersection. It analyzes the pressure on the road surface in real time, and combines the analyzed data to simulate the cracks and potholes in the actual intersection scene. Based on the constructed intersection scene, it predicts the current traffic flow at the intersection. According to the predicted traffic flow, it adjusts and manages the actual intersection in a timely manner to reduce traffic flow. It can effectively construct a crack scene of the road surface, improve the details of the intersection, better help people manage and assist traffic, and provide convenience. Attached Figure Description
[0015] Figure 1 A flowchart of the method framework for this invention is provided. Figure 2 A flowchart illustrating the steps involved in setting up the scenario for this invention; Figure 3 This is a flowchart illustrating the calculation of road surface stress conditions according to the present invention; Figure 4 This is a flowchart illustrating the process of predicting traffic flow at intersections according to the present invention. Figure 5 This is a block diagram of the system for building a hybrid scenario according to the present invention. Detailed Implementation
[0016] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.
[0017] Reference Figure 1 As shown, the method for building a hybrid digital scene based on AI artificial intelligence includes the following steps: Drones are used to scan and identify traffic intersections to obtain 3D point cloud data; The collected data is preprocessed and uploaded to modeling software to construct the intersection, which is then built into the computer. Obtain the current traffic flow at the intersection, and based on the traffic flow, determine the frequency of container truck appearances; Based on the frequency of the trucks, the stress on the road surface is calculated. Based on the stress, the corresponding road surface potholes and cracks are built on the road surface of the traffic intersection model. Based on the constructed digital intersection scenario, the traffic flow and road surface potholes at the intersection are simulated and predicted, and adjustments are made in the real environment.
[0018] This application predicts road surface damage, such as potholes and cracks, based on vehicle frequency and stress analysis, enabling timely maintenance and preventing traffic accidents and intersection congestion caused by road surface deterioration. This preventative maintenance can significantly extend the service life of the road surface and reduce maintenance costs. The prediction can help identify potential hazards, and based on the prediction results, traffic management departments can make scientific and reasonable decisions to manage traffic intersections, thus promoting the intelligent management of urban traffic.
[0019] Reference Figure 2 As shown, the drone is equipped with a 3D laser scanner. The drone flies above the intersection and uses the 3D laser scanner to scan the intersection. Based on the position of the intersection point between the laser beam and the object surface, it obtains 3D point cloud data of the intersection. The obtained data is preprocessed and then uploaded to the modeling software to construct the intersection. The 3D point cloud data preprocessing includes point cloud filtering, point cloud key point extraction, and point cloud registration.
[0020] The intersection model is constructed using the Poisson reconstruction algorithm. The processed data is then used in modeling software to construct an octree based on the density of the point cloud data. A function is defined for the octree, with the following form: Where X(q) is an indicator function used to define whether a point q in space is located on the surface of an object M, where q is a point in the three-dimensional point cloud data space, and M is the measured object; A vector field is constructed using point cloud data, the Poisson equation is solved using the vector field, isosurfaces are extracted, and reconstruction is completed. The formula for a vector field is: Where V is the normal vector of all P in the point cloud data. Let p be a point in the point cloud, and w be a point element. Let n be the normal vector of point P, and n be the neighborhood, which is the set of other points in the space surrounding point P. The extracted isosurfaces are integrated with the original point data to form a complete 3D intersection model.
[0021] In this application, the 3D point cloud data is a vector combination set in a 3D coordinate system, which has high flexibility and variable data density. It is scanned by a 3D laser scanner configured on a UAV. The 3D point cloud data processing method is existing technology and will not be elaborated here. The Poisson reconstruction algorithm is essentially an implicit function surface reconstruction algorithm. It uses a surface in space to distinguish between inside and outside, which can be intuitively understood as outside and inside the surface. It uses 0 and 1 to represent the inside and outside surfaces, that is, if an element belongs to this set, it is 1, and if an element does not belong to this set, it is 0. The gradient is calculated by using a function to calculate the gradient of all indicator functions inside the space. By solving the Poisson equation, the isosurface is extracted to obtain the surface of the object. Solving the Poisson equation using a vector field is existing technology and will not be elaborated here.
[0022] Reference Figure 3 As shown, historical traffic flow data of intersections is obtained based on intersection monitoring. Image analysis is used to obtain the traffic flow data of the monitored intersections. Within a unit of time, the number of times a container truck appears at the current intersection is counted. This data is used as the first set of data. Multiple sets of data samples are obtained as the dataset. The average frequency of container truck occurrences in the dataset is obtained. The frequency calculation formula is: Where L is the frequency of truck appearance, i is the number of times trucks appear, and t is the unit of time.
[0023] To calculate the pressure exerted on the road surface by a container truck, the weight of the fully loaded truck is obtained beforehand and estimated based on the truck's load limit. The calculation formula is as follows: F=G=mo Where G is gravity, m is mass, o is the speed of the truck traveling on the road, o is based on the average speed of the truck when it passes the road, and F is the pressure on the road surface. The formula for calculating the pressure on the road surface is: P is the pressure on the road surface, F is the force on the road surface, and S is the force-bearing area, where the force-bearing area is the area of the road surface that the tire rolls over. The contact surface between the tire and the ground is a rectangle or ellipse, and the force-bearing area is calculated based on the tire width and the predicted contact length.
[0024] The prediction of road surface cracks and potholes is based on factors including the road surface material structure coefficient, the frequency of trucks at intersections, and the construction quality coefficient. The road surface material structure coefficient is derived from road construction standards. The pothole assessment formula is as follows: Where R is the pressure coefficient of the road material, L is the frequency of truck occurrence, P is the pressure on the road surface, B is the construction quality coefficient, and H is the probability of potholes appearing on the road surface. The road surface texture effect is built using modeling software, including cracks and potholes.
[0025] Since there are many types and a wide range of classifications of container trucks involved in this application, and the situations they involve are all different, the container trucks involved in this application are uniformly conceived as trucks transporting construction materials. The parameters of the container trucks can be changed according to the actual situation. The image analysis method is existing technology and will not be elaborated on here. The speed of the container trucks traveling on the road is obtained based on traffic monitoring. The speed of the container trucks when passing through intersections is less than 15km / h. By accurately simulating cracks and potholes on real road surfaces, the complexity and diversity of actual roads can be highly restored, making the simulated environment closer to the real scene. The realistic road surface effect can enhance the visual experience of the observer. In the road maintenance stage, by simulating the development process of road surface defects, the evolution trend of defects can be predicted, providing a scientific basis for formulating reasonable maintenance plans and improving maintenance efficiency and effectiveness.
[0026] Using deep learning algorithms in computers, based on the obtained road surface pothole probabilities, we construct the effects of cracks and potholes on traffic roads, and verify the constructed crack and pothole effects to see if they match the actual scenario. Based on task requirements, a deep learning network architecture is selected, and the model is trained using training and validation sets. During training, changes in the loss function and accuracy metrics are monitored, and the model's performance is evaluated using a test set. GANs are used to generate images similar to real cracks and potholes to further expand the training dataset. GANs consist of a generator and a discriminator, and high-quality images are generated through adversarial training. The obtained road surface pothole probabilities are simulated to construct a dataset of road surface crack and pothole effects under different scenarios. The validation formula is as follows: (This is a partial translation of the original text, and the translation is not possible without the full context.) Where Z is a given feature Under these conditions, does the dataset input from the road surface belong to the normal crack analogy? This represents the distance threshold between cracks and potholes on the road and the roadside in a real-world scenario. The parameter vector of the model is input with the features of the constructed road surface crack and pothole effect dataset to determine whether the constructed road surface effect matches the actual situation.
[0027] In this application, the road surface crack and pothole effects are randomly generated. Since vehicles travel in the middle of the road in the motor vehicle lane, while the two sides of the road are for non-motorized vehicles, the probability of cracks and potholes appearing on the sides of the road is extremely low and is not considered in this application. The distance threshold of the roadside is the width value of the non-motorized vehicle lanes on both sides. The randomly generated road surface crack and pothole effects are verified to see if they conform to the actual road surface conditions.
[0028] Reference Figure 4 As shown, a traffic flow prediction model is constructed. Based on real-time acquired traffic flow data, the data is analyzed. The formula for calculating traffic flow is as follows: U=A*D Where U represents the current traffic flow data at the intersection, A represents the traffic flow at the intersection under natural conditions, which is the total number of vehicles passing through the intersection per unit time, and D represents the external influencing factors. External influencing factors include morning and evening rush hours, weather conditions, accident occurrences, road construction at intersections, and potholes on the road surface. Traffic volume is higher than average during morning and evening rush hours, while traffic volume increases during rainy or snowy days. Accidents, road construction at intersections, and potholes on the road surface cause a decrease in traffic volume. The formula for calculating the external influencing factors is as follows: D=J 时间 *J 天气 *J 事故 *J 施工 *H J 时间 J is the time factor. 天气 As a weather factor, J 事故 As an accident factor, J 施工 H represents the construction factor, and H represents the road surface probability. The obtained traffic flow dataset is substituted into the model for validation.
[0029] Based on the constructed traffic flow prediction model, traffic flow is simulated in a digital scenario. Based on the simulation results, traffic flow at the current intersection is controlled and adjusted. Traffic flow is directly proportional to the occurrence of accidents; the more vehicles there are, the more likely traffic accidents are to occur. During periods of high traffic flow, traffic controllers are deployed to direct traffic and reduce the probability of accidents. During periods of low traffic flow, road surface construction is carried out.
[0030] This application utilizes a traffic flow prediction model to forecast traffic flow in real time over a future period. During periods when a significant increase in traffic flow is predicted, traffic management departments can take preventative measures, such as adjusting traffic light timings, lane allocation, and setting up temporary traffic signs, to alleviate traffic congestion and reduce the probability of traffic accidents. Based on the traffic flow prediction results, traffic management departments can rationally allocate police resources to ensure sufficient manpower for road patrols and emergency response during peak traffic hours. Through scientific traffic control measures, traffic light timings and lane allocations at intersections can be optimized, improving road efficiency and reducing vehicle waiting time, thereby alleviating traffic congestion. Factors affecting road conditions include traffic load, environmental factors, material quality, and design and construction. Traffic load is the most significant factor influencing road quality. Other influencing factors have been considered and measured during road planning, design, and construction, and their impact is relatively low; therefore, they are not considered in this paper. However, the impact of truck traffic cannot be measured during road operation; frequent truck traffic makes roads more susceptible to damage.
[0031] Reference Figure 5 As shown, the AI-based digital scene hybrid construction system includes a modeling module, a road surface stress analysis module, a prediction module, and a data acquisition module. The modeling module uses a computer as a carrier to build the hybrid scene, the road surface stress analysis module analyzes the stress on the road surface, the prediction module predicts and analyzes the actual intersection conditions based on the constructed digital scene, and the data acquisition module collects data from the actual intersection to construct a dataset.
[0032] The data acquisition module uses drones to collect real-time data on the traffic intersection and obtains 3D point cloud data of the intersection. The module also obtains image data of the intersection based on the intersection's monitoring and processes the acquired images using image analysis methods.
[0033] This application's data acquisition module utilizes drones to collect real-time data on traffic intersection conditions, enabling rapid acquisition of 3D point cloud data. This acquisition method is not only efficient but also covers a wider area, capturing more details and providing rich data support for subsequent modeling and prediction. It also acquires image data from intersection monitoring equipment and processes the images based on image analysis technology. This approach reflects the traffic conditions at the intersection in real time, providing intuitive and accurate data for traffic management. The collected 3D point cloud data and monitoring image data complement each other, forming a multi-dimensional description of the intersection conditions. This data combination method improves the accuracy and comprehensiveness of the data. The road surface stress analysis module performs in-depth analysis of the road surface stress, assessing the road's load-bearing capacity and stability, and enabling rapid data acquisition, processing, and analysis.
[0034] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.
Claims
1. A method for hybrid construction of real-world digital scenes based on AI artificial intelligence, characterized in that: The setup methods include: Drones are used to scan and identify traffic intersections to obtain 3D point cloud data; The collected data is preprocessed and uploaded to modeling software to construct the intersection, which is then built into the computer. Obtain the current traffic flow at the intersection, and based on the traffic flow, determine the frequency of container truck appearances; Based on the frequency of the trucks, the stress on the road surface is calculated. Based on the stress, the corresponding road surface potholes and cracks are built on the road surface of the traffic intersection model. Based on the constructed digital intersection scenario, the traffic flow and road surface potholes at the intersection are simulated and predicted, and adjustments are made in the real environment. The formula for assessing road surface potholes is: ； Where R is the pressure coefficient of the road material, L is the frequency of truck occurrence, P is the pressure on the road surface, B is the construction quality coefficient, and H is the probability of potholes appearing on the road surface. The road surface texture effect is built using modeling software, including cracks and potholes; Using deep learning algorithms in computers, based on the obtained road surface pothole probabilities, we construct the effects of cracks and potholes on traffic roads, and verify the constructed crack and pothole effects to see if they match the actual scenario. The validation formula is as follows: (This is a partial translation of the original text, and the translation is not possible without the full context.) ； Where Z is a given feature Under these conditions, does the dataset input from the road surface belong to the normal crack analogy? This represents the distance threshold between cracks and potholes on the road and the roadside in a real-world scenario. The parameter vector of the model is input with the features of the constructed road surface crack and pothole effect dataset to determine whether the constructed road surface effect matches the actual situation.
2. The method for hybrid construction of AI-based real-world digital scenes according to claim 1, characterized in that, The drone is equipped with a 3D laser scanner. The drone flies over the intersection and uses the 3D laser scanner to scan the intersection. Based on the position of the intersection point between the laser beam and the object surface, it acquires 3D point cloud data of the intersection. The acquired data is preprocessed and then uploaded to modeling software to construct the intersection. The 3D point cloud data preprocessing includes point cloud filtering, point cloud key point extraction, and point cloud registration.
3. The method for hybrid construction of AI-based real-world digital scenes according to claim 2, characterized in that, The intersection model is constructed using the Poisson reconstruction algorithm. The processed data is then used in modeling software to construct an octree based on the density of the point cloud data. A function is defined for the octree, with the following form: ； Where X(q) is an indicator function used to define whether a point q in space is located on the surface of an object M, where q is a point in the three-dimensional point cloud data space, and M is the measured object; A vector field is constructed using point cloud data, the Poisson equation is solved using the vector field, isosurfaces are extracted, and reconstruction is completed. The formula for a vector field is: ； Where V is the normal vector of all P in the point cloud data. Let p be a point in the point cloud, and w be a point element. Let n be the normal vector of point P, and n be the neighborhood, which is the set of other points in the space surrounding point P. The extracted isosurfaces are integrated with the original point data to form a complete 3D intersection model.
4. The method for hybrid construction of AI-based real-world digital scenes according to claim 1, characterized in that, Historical traffic flow data from intersections is obtained through intersection monitoring. Image analysis is used to obtain the traffic flow data from the monitored intersections. Within a unit of time, the number of times a container truck appears at the current intersection is counted. This data is used as the first set of data. Multiple data samples are obtained to form the dataset. The average frequency of container truck appearances in the dataset is calculated using the following formula: ； Where L is the frequency of truck appearance, i is the number of times trucks appear, and t is the unit of time.
5. The method for hybrid construction of AI-based real-world digital scenes according to claim 4, characterized in that, To calculate the pressure exerted on the road surface by a container truck, the weight of the fully loaded truck is obtained beforehand and estimated based on the truck's load limit. The calculation formula is as follows: F=G=mo Where G is gravity, m is mass, o is the speed of the truck traveling on the road, o is based on the average speed of the truck when it passes the road, and F is the pressure on the road surface. The formula for calculating the pressure on the road surface is: ； P is the pressure on the road surface, F is the force on the road surface, and S is the force-bearing area, where the force-bearing area is the area of the road surface that the tire rolls over. The contact surface between the tire and the ground is a rectangle or ellipse, and the force-bearing area is calculated based on the tire width and the predicted contact length.
6. The method for hybrid construction of AI-based real-world digital scenes according to claim 5, characterized in that, The prediction of road surface cracks and potholes is based on factors including the road surface material structure coefficient, the frequency of trucks at intersections, and the construction quality coefficient. The road surface material structure coefficient is derived from road construction standards. The pothole assessment formula is as follows: ； Where R is the pressure coefficient of the road material, L is the frequency of truck occurrence, P is the pressure on the road surface, B is the construction quality coefficient, and H is the probability of potholes appearing on the road surface. The road surface texture effect is built using modeling software, including cracks and potholes.
7. A system for building a hybrid digital scene based on AI, characterized in that: The scene hybrid construction system includes: a modeling module, a road surface stress analysis module, a prediction module, and a data acquisition module. The modeling module uses a computer as a carrier to build hybrid scenes. The road surface stress analysis module analyzes the stress on the road surface. The prediction module predicts and analyzes the actual intersection conditions based on the constructed digital scene. The data acquisition module collects data from actual intersections and constructs a dataset.
8. The AI-based artificial intelligence-driven hybrid digital scene construction system according to claim 7, characterized in that, The data acquisition module uses drones to collect real-time data on the traffic intersection and obtains 3D point cloud data of the intersection. The module also obtains image data of the intersection based on the intersection's monitoring and processes the acquired images using image analysis methods.