Adjustment method for predicting double-domain physical collision situation and terminal
By generating a dual-domain object set in AR/MR technology and using a positional dynamics solver for iterative calculation and threshold adjustment, the problem of lack of closed-loop calibration in existing technologies is solved, achieving high-precision and stable virtual reality collision detection.
Patent Information
- Application Number
- CN202511448970.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-11
- Publication Date
- 2026-02-24
AI Technical Summary
The lack of a closed-loop calibration mechanism between prediction results and actual observation data in existing AR/MR technologies leads to virtual objects penetrating or false alarms, affecting user experience and system stability.
By generating models and bounding box parameters in the real and virtual domains, merging them into a dual-domain object set, and using a positional dynamics solver for iterative calculation to generate an initial detection threshold, the threshold is then adjusted using the error vector to achieve closed-loop calibration between prediction and actual measurement.
It achieves predictive detection and adaptive adjustment for collisions between two-domain objects, reducing missed detections and false alarms, and improving the accuracy and stability of detection.
Smart Images
Figure CN121564285A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of AR technology, and in particular to an adjustment method and terminal for predicting dual-domain physical collision scenarios. Background Technology
[0002] With the development of augmented reality (AR) and mixed reality (MR) technologies, the demand for collision detection between virtual objects and the real environment is increasing. Current technologies typically employ two types of methods for collision detection: one relies on the physical calculations built into the virtual engine, determining collisions only between objects in the virtual domain, and failing to adequately consider the dynamic information of the real environment; the other is based on point cloud data collected by sensors, detecting potential collision events in the real domain through spatial reconstruction. While both methods can achieve collision detection to some extent, they both have significant limitations.
[0003] One key shortcoming is the lack of a closed-loop calibration mechanism between predicted results and actual observation data in existing technologies. Current methods often set a fixed threshold to determine risk during the prediction phase. However, under different scenarios and object conditions, this threshold may be too lenient, leading to missed detections, or too strict, leading to false alarms. Once there is a deviation between the predicted result and the actual measurement, the system cannot dynamically correct the error, causing the detection accuracy to remain at a fixed level for a long time. This deficiency is particularly pronounced in mobile AR / MR applications, where the environment is complex and computing power is limited, making it prone to virtual objects penetrating real-world objects or false alarms, severely impacting user experience and system stability. Summary of the Invention
[0004] The technical problem to be solved by the present invention is to provide an adjustment method and terminal for predicting dual-domain physical collision scenarios, which can realize predictive detection and adaptive adjustment of dual-domain object collision scenarios.
[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: An adjustment method for predicting dual-domain physical collision scenarios includes the following steps: S1. Obtain the real domain model and its corresponding bounding box parameters, and the virtual domain model and its corresponding bounding box parameters; S2. Merge the real domain model and its bounding box parameters and the virtual domain model and its bounding box parameters to generate a dual-domain object set; S3. Iteratively calculate the collision scenarios of the dual-domain object set using a positional dynamics solver to generate a predicted state vector; generate an initial detection threshold based on the bounding box parameters of the real domain model and the virtual domain model; the initial detection threshold is used to determine the risk of collision scenarios of the dual-domain object set. S4. Obtain the actual state vector generated by the collision between the real domain model and the virtual domain model, compare the actual state vector with the predicted state vector and generate an error vector; adjust the initial detection threshold according to the error vector.
[0006] To solve the above-mentioned technical problems, another technical solution adopted by the present invention is as follows: An adjustment terminal for predicting dual-domain physical collision scenarios includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it performs the following steps: S1. Obtain the real domain model and its corresponding bounding box parameters, and the virtual domain model and its corresponding bounding box parameters; S2. Merge the real domain model and its bounding box parameters and the virtual domain model and its bounding box parameters to generate a dual-domain object set; S3. Iteratively calculate the collision scenarios of the dual-domain object set using a positional dynamics solver to generate a predicted state vector; generate an initial detection threshold based on the bounding box parameters of the real domain model and the virtual domain model; the initial detection threshold is used to determine the risk of collision scenarios of the dual-domain object set. S4. Obtain the actual state vector generated by the collision between the real domain model and the virtual domain model, compare the actual state vector with the predicted state vector and generate an error vector; adjust the initial detection threshold according to the error vector.
[0007] The beneficial effects of this invention are as follows: It provides an adjustment method and terminal for predicting dual-domain physical collision scenarios. First, models and bounding box parameters are generated in the real domain and virtual domain respectively, enabling objects in both domains to be represented under a unified data structure. Then, the two are merged to form a dual-domain object set, which is iteratively solved using a positional dynamics solver. An initial detection threshold is generated based on the bounding box parameters to establish a safety benchmark adapted to the object's geometry for risk assessment. An error vector is then generated by comparing the predicted state vector and the actual state vector, and this error vector is used to dynamically adjust the detection threshold, thereby achieving closed-loop calibration between prediction and measurement. Thus, predictive detection and adaptive adjustment of dual-domain object collision scenarios are achieved, reducing missed detections and false alarms, and improving the accuracy and stability of detection. Attached Figure Description
[0008] Figure 1 This is a flowchart illustrating an adjustment method for predicting dual-domain physical collision scenarios according to an embodiment of the present invention. Figure 2 This is a detailed flowchart illustrating an adjustment method for predicting dual-domain physical collision scenarios according to an embodiment of the present invention. Figure 3 This is a schematic diagram of an adjustment terminal for predicting dual-domain physical collision scenarios according to an embodiment of the present invention. Label Explanation: 1. An adjustment terminal for predicting dual-domain physical collision scenarios; 2. A memory; 3. A processor. Detailed Implementation
[0009] To explain in detail the technical content, objectives, and effects of the present invention, the following description is provided in conjunction with the embodiments and accompanying drawings.
[0010] Before detailing the embodiments of this application, some related concepts will first be explained: The reality domain refers to objects and their geometric data existing in the real physical environment, collected by sensors, such as point cloud data of tables, walls, and players' arms obtained by depth cameras or LiDAR. After semantic segmentation and abstract modeling, these real-world objects are transformed into reality domain models, which, along with corresponding geometric parameters, participate in subsequent collision prediction and verification.
[0011] A virtual domain refers to a collection of virtual objects generated by a graphics engine or simulation system. These objects have preset geometric shapes, motion trajectories, and material properties, such as virtual bullets, virtual characters, or game props. During operation, the state information of virtual domain objects can be acquired in real time and participate in physics solving together with real domain objects, thereby achieving collision detection under virtual-real fusion.
[0012] Bounding box parameters are important indicators for unifying the geometric abstraction of real-world and virtual-world objects. These parameters include the object's length, width, height, centroid position, diagonal length, and the maximum radius from the centroid to the surface. By approximating complex objects with simplified geometry, bounding boxes not only significantly reduce computational complexity but also provide a unified quantitative basis for calculating the minimum surface spacing between objects, setting safe distance thresholds, and converting angular velocities.
[0013] Position-based dynamics solvers are a physics computation method based on iterative constraints. Their principle lies in directly satisfying geometric constraints through position adjustments, rather than relying on traditional differential equation solutions. In this invention, the position-based dynamics solver is used to iteratively compute a set of dual-domain objects within a short prediction time window. Combined with continuous collision detection constraints, it prevents high-speed or rotating objects from tunneling through the object, thereby efficiently and stably generating predicted state vectors that meet the dual requirements of real-time performance and accuracy for mobile devices.
[0014] In existing technologies, with the rapid development of augmented reality (AR) and mixed reality (MR) applications, the demand for interaction between virtual objects and the real environment is constantly increasing, making collision detection a core element in ensuring the realism and safety of the interaction. Traditional methods often rely on single-domain data processing, such as using only physical calculations in the virtual engine to simulate collisions, or relying solely on real-world point cloud data collected by sensors for detection. These approaches have significant shortcomings: firstly, virtual domain computation ignores the dynamic changes in the real environment, easily leading to "clipping" or "missed collisions"; secondly, the accuracy of point cloud detection in the real domain is limited by device computing power and sampling accuracy, making it difficult to balance real-time performance and stability. Furthermore, most existing detection methods are performed offline or in one-way data transmission modes, lacking the ability to predict future time windows and a closed-loop verification mechanism between predicted results and actual observation data. Especially in mobile terminal environments, limited computing power and latency sensitivity make it impossible to achieve stable and reliable virtual-real fusion collision detection in complex scenes. These shortcomings result in low testing efficiency and high false positive rates, making it difficult to meet the demands of AR / MR games, industrial simulations, and security monitoring scenarios for high real-time performance, high accuracy, and adaptability.
[0015] To at least solve the above problems, please refer to Figure 1 This invention provides an adjustment method for predicting dual-domain physical collision scenarios, comprising the following steps: S1. Obtain the real domain model and its corresponding bounding box parameters, and the virtual domain model and its corresponding bounding box parameters; S2. Merge the real domain model and its bounding box parameters and the virtual domain model and its bounding box parameters to generate a dual-domain object set; S3. Iteratively calculate the collision scenarios of the dual-domain object set using a positional dynamics solver to generate a predicted state vector; generate an initial detection threshold based on the bounding box parameters of the real domain model and the virtual domain model; the initial detection threshold is used to determine the risk of collision scenarios of the dual-domain object set. S4. Obtain the actual state vector generated by the collision between the real domain model and the virtual domain model, compare the actual state vector with the predicted state vector and generate an error vector; adjust the initial detection threshold according to the error vector.
[0016] As described above, the beneficial effects of this invention are as follows: First, models and bounding box parameters are generated in the real domain and virtual domain respectively, enabling objects in both domains to be represented under a unified data structure. Then, the two are merged to form a dual-domain object set, and continuous collision detection constraints are introduced into the positional dynamics solver for iterative solving of this set. By generating an initial detection threshold based on the bounding box parameters, a safety benchmark adapted to the object's geometry can be established for risk assessment. Then, an error vector is generated by comparing the predicted state vector and the actual state vector, and this error vector is used to dynamically adjust the detection threshold, thereby achieving closed-loop calibration between prediction and measurement. Thus, the principle of this scheme lies in the process of "unified modeling—prediction and solution—threshold determination—error calibration," ultimately achieving predictive detection and adaptive adjustment of collision scenarios for dual-domain objects, reducing missed detections and false alarms, and improving the accuracy and stability of detection.
[0017] In some implementations, step S1 specifically includes: S11. Obtain the geometric outline of the real-domain object and calculate and generate a real-domain model based on the geometric outline; configure and generate bounding box parameters based on the real-domain model; S12. Obtain the geometric parameters of the virtual domain object and generate a virtual domain model based on the geometric parameters; configure and generate bounding box parameters based on the virtual domain model.
[0018] As described above, the real-domain object is calculated to obtain the real-domain model, and bounding box parameters are generated; the virtual-domain object is generated to produce a virtual-domain model based on geometric parameters, and bounding box parameters are also configured. The principle behind this design is to abstract both real-domain and virtual-domain objects into a unified form of "model + bounding box," ensuring consistency in subsequent merging and solving. Since the real-domain is mostly derived from point clouds or sensor reconstruction, while the virtual domain originates from engine parameters, without unified abstraction, inconsistencies in inputs and deviations in calculation results are easily caused. This approach ensures the comparability and uniformity of the dual-domain models in terms of spatial extent, centroid location, and scale parameters. Its effect is to reduce errors caused by different sources of geometric data, improve the accuracy of subsequent iterative predictions, and provide a reliable geometric benchmark for threshold determination.
[0019] In some embodiments, step S11 further includes: Obtain point cloud data of real-world objects, perform semantic segmentation of the point cloud data according to preset categories, and generate a point cloud set; The level of detail is calculated based on the point cloud set, and the real-world object is abstractly modeled according to the level of detail to generate a real-world model.
[0020] As described above, point cloud data processing and level of detail control are further introduced during the generation of the real-domain model. The specific principle is as follows: through semantic segmentation, the point cloud is divided into static objects, dynamic objects, player characters, and negligible objects, thereby filtering out irrelevant information and retaining key objects. Then, the level of detail for modeling is determined by calculating the point cloud density: high-precision modeling is used when the density is high, medium-precision modeling is used when the density is moderate, and low-precision modeling is used when the density is low. This adaptively balances modeling accuracy and computational overhead. The effect is that the system can operate stably under different hardware conditions: high-precision detection can be maintained on high-end devices, while real-time performance and basic reliability can still be guaranteed on low-end devices. Ultimately, this achieves cross-device applicability and rational utilization of computational resources.
[0021] In some implementations, step S3 specifically includes: S31. Obtain the collision speed parameters and maximum outer radius of the dual-domain object set, and construct continuous collision detection constraints based on the collision speed parameters and the maximum outer radius; S32. Based on the continuous collision detection constraint, the positional dynamics solver is used to iteratively calculate the dual-domain object set to obtain the predicted state vector for a future preset time period and the minimum surface spacing of the dual-domain object set. S33. An initial detection threshold is generated based on the minimum surface spacing of the dual-domain object set and the bounding box parameters. The initial detection threshold is used to determine the risk of collision between dual-domain objects.
[0022] As described above, in the prediction phase, the system constructs continuous collision detection constraints based on the velocity and maximum circumscribed radius of the dual-domain objects. The principle is to convert linear and angular velocities into geometric expansion using the expansion distance formula, thereby preventing high-speed or rotating objects from "tunneling" within the prediction step. Next, a positional dynamics solver iteratively calculates the dual-domain object set to obtain the predicted state vector and minimum surface spacing for a preset future time period. This is then combined with bounding box parameters (such as diagonal length) to generate an initial detection threshold, which serves as the benchmark for risk assessment. When the predicted spacing is less than the threshold, a potential risk is considered; otherwise, it is considered safe. The effect is that it not only predicts collision scenarios in the future time period but also dynamically adjusts the risk assessment threshold based on object size, improving detection sensitivity and avoiding biased judgments of large or small objects, thus enhancing the rationality and practicality of the early warning system.
[0023] In some implementations, step S4 specifically includes: S41. Obtain the actual state vector generated by the two-domain collision; S42. Compare the actual state vector with the predicted state vector to obtain the position deviation and velocity deviation; S43. Generate an error vector based on the position deviation and the velocity deviation; S44. Based on the error vector, the initial detection threshold is reduced or increased to obtain a corrected detection threshold, and the corrected detection threshold is used to determine the risk of collision between the two-domain objects.
[0024] As described above, during the verification phase, the system collects the actual state vector of the two-domain collision and compares it item by item with the predicted state vector to obtain the position deviation and velocity deviation, thereby generating an error vector. The principle lies in quantifying the difference between the predicted and measured data by comparing them, and using the error vector to tighten or loosen the initial detection threshold. Specifically, when the error is small for several consecutive frames, it indicates reliable prediction, and the threshold can be tightened to improve sensitivity; when the error is large for several consecutive frames, it indicates unstable prediction, and the threshold should be loosened to reduce the false alarm rate. The final corrected detection threshold is then used for risk assessment. Its effect is to construct a dynamic closed-loop system: the system can automatically adjust the judgment criteria based on the prediction accuracy, thereby maintaining the stability and robustness of the detection performance. Through this process, it can ensure sensitive early warning under high prediction accuracy and reduce erroneous judgments in complex scenarios, ultimately improving the method's adaptability and long-term stability.
[0025] An adjustment terminal 1 for predicting dual-domain physical collision scenarios includes a memory 2, a processor 3, and a computer program stored in the memory 2 and running on the processor 3. When the processor 3 executes the computer program, it implements the steps in an adjustment method for predicting dual-domain physical collision scenarios.
[0026] Please refer to Figure 1 and Figure 2 One embodiment of the present invention is as follows: In one specific embodiment, the present invention provides an adjustment method for predicting dual-domain physical collision scenarios. The method includes the following steps: S11. The system acquires point cloud data of real-world objects using a depth camera and performs semantic segmentation on the point cloud data according to preset categories, classifying them into static objects, dynamic objects, player characters, and negligible objects, thus forming a point cloud set. Based on the point cloud set, the point density (LOD level) is calculated. When the point density is high, high-precision modeling is performed to retain more geometric details; when the point density is within a normal range, medium-precision modeling is performed to balance accuracy and computational power; and when the point density is low, low-precision modeling is performed to ensure real-time performance. A real-world model is generated in this way. Subsequently, the outline of the real-world model is extracted, and its length, width, height, centroid position, diagonal length, and maximum circumscribed radius are calculated to form the bounding box parameters of the real-world objects.
[0027] S12. The virtual domain object is provided with geometric and motion parameters by the engine. A virtual domain model is generated based on its geometric information, and corresponding bounding box parameters are configured. This bounding box is updated in real time when the object rotates or scales to ensure parameter accuracy. Through the above process, both the real domain and virtual domain objects possess a unified abstract expression of "model + bounding box"; the bounding box parameters for both domains include length, width, height, centroid position, and diagonal length. S2. Unify the real-domain model and its bounding box parameters with the virtual-domain model and its bounding box parameters into the same coordinate system, merging them to form a dual-domain object set. This set provides the input basis for subsequent unified dynamics solutions, ensuring the comparability of real-domain and virtual-domain data in terms of spatial range and scale.
[0028] S31. Obtain the collision velocity parameters and maximum outer radius of the two-domain object collection, according to the formula:
[0029] In the formula, Indicates the expansion distance; Indicates the linear velocity of the collision; Indicates the angular velocity of the collision; This represents the distance from the center of mass to the farthest surface of the bounding box; Indicates the prediction step size; The expansion distance is calculated and converted into a continuous collision detection constraint (CCD constraint) to prevent high-speed or rotating objects from tunneling through the prediction time window.
[0030] S32. Using a positional dynamics solver, iterative calculations are performed on the set of dual-domain objects within a set prediction time window to obtain the predicted state vector for future moments, including position, linear velocity, and angular velocity information, and the minimum surface distance between the dual-domain objects is calculated.
[0031] S33. Subsequently, based on the diagonal length d in the bounding box parameters, an initial detection threshold is generated, for example... This threshold serves as the risk assessment benchmark: if the predicted minimum distance is less than δ, a potential collision risk is identified; if it is greater than or equal to δ, the situation is considered safe. Thus, the initial detection threshold is directly related to the object size and spatial distance, ensuring the rationality and adaptability of the risk assessment.
[0032] S41. After the prediction time window ends, the system collects the actual state vector of the real domain object obtained by the sensor and combines it with the synchronous state vector of the virtual domain object to form the actual state vector.
[0033] S42. Compare the actual state vector with the predicted state vector item by item to obtain the position deviation. Speed deviation And the velocity deviation calculated from angular velocity. .
[0034] S43. Combine these deviations to generate an error vector, and calculate the error norm as a quantitative indicator of the overall prediction error.
[0035] S44. Adjust the initial detection threshold based on the error vector: When the error norm is less than the threshold for several consecutive frames, it indicates that the prediction accuracy is high. At this time, tighten the threshold to improve sensitivity; when the error norm is greater than the threshold for several consecutive frames, it indicates that there is a prediction deviation. At this time, relax the threshold to reduce the false alarm rate. The corrected detection threshold is obtained from this, and the risk assessment of the prediction state is performed again using the corrected threshold, realizing a closed-loop process of prediction-measurement-adjustment.
[0036] Please refer to Figure 3 Another embodiment of the present invention is as follows: An adjustment terminal 1 for predicting dual-domain physical collision scenarios includes a memory 2, a processor 3, and a computer program stored in the memory 2 and running on the processor 3. When the processor 3 executes the computer program, it implements the steps in an adjustment method for predicting dual-domain physical collision scenarios.
[0037] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent modifications made based on the content of the present invention specification and drawings, or direct or indirect applications in related technical fields, are similarly included within the patent protection scope of the present invention.
Claims
1. An adjustment method for predicting dual-domain physical collision scenarios, characterized in that, Including the following steps: S1. Obtain the real domain model and its corresponding bounding box parameters, and the virtual domain model and its corresponding bounding box parameters; S2. Merge the real domain model and its bounding box parameters and the virtual domain model and its bounding box parameters to generate a dual-domain object set; S3. Iteratively calculate the collision scenarios of the dual-domain object set using a positional dynamics solver to generate a predicted state vector; generate an initial detection threshold based on the bounding box parameters of the real-domain model and the virtual-domain model; The initial detection threshold is used to determine the risk of collisions in the dual-domain object set; S4. Obtain the actual state vector generated by the collision between the real domain model and the virtual domain model, compare the actual state vector with the predicted state vector and generate an error vector; adjust the initial detection threshold according to the error vector.
2. The adjustment method for predicting dual-domain physical collision scenarios according to claim 1, characterized in that, Step S1 specifically includes: S11. Obtain the geometric outline of the real-domain object and calculate and generate a real-domain model based on the geometric outline; configure and generate bounding box parameters based on the real-domain model; S12. Obtain the geometric parameters of the virtual domain object and generate a virtual domain model based on the geometric parameters; configure and generate bounding box parameters based on the virtual domain model.
3. The adjustment method for predicting dual-domain physical collision scenarios according to claim 2, characterized in that, Step S11 further includes: Obtain point cloud data of real-world objects, perform semantic segmentation of the point cloud data according to preset categories, and generate a point cloud set; The level of detail is calculated based on the point cloud set, and the real-world object is abstractly modeled according to the level of detail to generate a real-world model.
4. The adjustment method for predicting dual-domain physical collision scenarios according to claim 1, characterized in that, Step S3 specifically includes: S31. Obtain the collision speed parameters and maximum outer radius of the dual-domain object set, and construct continuous collision detection constraints based on the collision speed parameters and the maximum outer radius; S32. Based on the continuous collision detection constraint, the positional dynamics solver is used to iteratively calculate the dual-domain object set to obtain the predicted state vector for a future preset time period and the minimum surface spacing of the dual-domain object set. S33. An initial detection threshold is generated based on the minimum surface spacing of the dual-domain object set and the bounding box parameters. The initial detection threshold is used to determine the risk of collision between dual-domain objects.
5. The adjustment method for predicting dual-domain physical collision scenarios according to claim 1, characterized in that, Step S4 specifically includes: S41. Obtain the actual state vector generated by the two-domain collision; S42. Compare the actual state vector with the predicted state vector to obtain the position deviation and velocity deviation; S43. Generate an error vector based on the position deviation and the velocity deviation; S44. Based on the error vector, the initial detection threshold is reduced or increased to obtain a corrected detection threshold, and the corrected detection threshold is used to determine the risk of collision between the two-domain objects.
6. An adjustment terminal for predicting dual-domain physical collision scenarios, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it performs the following steps: S1. Obtain the real domain model and its corresponding bounding box parameters, and the virtual domain model and its corresponding bounding box parameters; S2. Merge the real domain model and its bounding box parameters and the virtual domain model and its bounding box parameters to generate a dual-domain object set; S3. Iteratively calculate the collision scenarios of the dual-domain object set using a positional dynamics solver to generate a predicted state vector; generate an initial detection threshold based on the bounding box parameters of the real-domain model and the virtual-domain model; The initial detection threshold is used to determine the risk of collisions in the dual-domain object set; S4. Obtain the actual state vector generated by the collision between the real domain model and the virtual domain model, compare the actual state vector with the predicted state vector and generate an error vector; adjust the initial detection threshold according to the error vector.
7. The adjustment terminal for predicting dual-domain physical collision scenarios according to claim 6, characterized in that, Step S1 specifically includes: S11. Obtain the geometric outline of the real-domain object and calculate and generate a real-domain model based on the geometric outline; configure and generate bounding box parameters based on the real-domain model; S12. Obtain the geometric parameters of the virtual domain object and generate a virtual domain model based on the geometric parameters; configure and generate bounding box parameters based on the virtual domain model.
8. The adjustment terminal for predicting dual-domain physical collision scenarios according to claim 7, characterized in that, Step S11 further includes: Obtain point cloud data of real-world objects, perform semantic segmentation of the point cloud data according to preset categories, and generate a point cloud set; The level of detail is calculated based on the point cloud set, and the real-world object is abstractly modeled according to the level of detail to generate a real-world model.
9. The adjustment terminal for predicting dual-domain physical collision scenarios according to claim 6, characterized in that, Step S3 specifically includes: S31. Obtain the collision speed parameters and maximum outer radius of the dual-domain object set, and construct continuous collision detection constraints based on the collision speed parameters and the maximum outer radius; S32. Based on the continuous collision detection constraint, the positional dynamics solver is used to iteratively calculate the dual-domain object set to obtain the predicted state vector for a future preset time period and the minimum surface spacing of the dual-domain object set. S33. An initial detection threshold is generated based on the minimum surface spacing of the dual-domain object set and the bounding box parameters. The initial detection threshold is used to determine the risk of collision between dual-domain objects.
10. The adjustment terminal for predicting dual-domain physical collision scenarios according to claim 6, characterized in that, Step S4 specifically includes: S41. Obtain the actual state vector generated by the two-domain collision; S42. Compare the actual state vector with the predicted state vector to obtain the position deviation and velocity deviation; S43. Generate an error vector based on the position deviation and the velocity deviation; S44. Based on the error vector, the initial detection threshold is reduced or increased to obtain a corrected detection threshold, and the corrected detection threshold is used to determine the risk of collision between the two-domain objects.