Semantic information acquisition method and system for building model

Abstract

The invention provides a semantic information acquisition method and system for a building model, and the method comprises the steps: importing a three-dimensional model of a to-be-constructed region, and selecting a current wall surface in the model; calculating the geometric contour of the current wall surface; judging whether a hallway exists in the geometric contour of the current wall surface or not; if yes, updating the geometric contour and semantic information of the current wall surface according to the boundary of the hallway; if not, maintaining the semantic information of the current wall surface; on the basis of the geometric contour of the current wall surface, calculating wall-to-wall distance information of the current wall surface, and supplementing the wall-to-wall distance information into semantic information of the current wall surface; and based on the geometric contour of the current wall surface, identifying holes, cross beams and bay windows of the current wall surface, and supplementing the holes, the cross beams and the bay windows into the semantic information of the current wall surface. According to the method, multi-dimensional collaborative operation is carried out, information such as the geometric contour, the spatial size, the local environment characteristics and the holes of the wall surface is comprehensively analyzed, so that a detailed semantic map is generated, rich environment information is provided for the building robot, and planning and execution of the building robot in complex tasks are supported.

Description

Technical Field

[0001] The present invention relates to the technical field of indoor task planning for construction robots, and in particular to a method and system for acquiring semantic information of a building model. Background Art

[0002] In indoor task planning for construction robots, accurately acquiring the semantic information of indoor building 3D models is the key to achieving efficient task planning. Existing methods for acquiring semantic information of indoor building 3D models in indoor task planning for construction robots have the following major drawbacks:

[0003] 1. Poor interactivity: Existing semantic information acquisition methods advocate outputting all geometric information at once, making it impossible to dynamically filter areas requiring planning based on specific on-site conditions.

[0004] 2. Low efficiency: Methods based on vision and depth information have high training costs, and there is no mature and large construction model dataset as a sample training environment. In addition, the computational complexity and data processing are high, which cannot meet the needs of real-time task planning.

[0005] It can be seen that existing technologies cannot meet the real-time task planning requirements of construction robots in complex indoor environments. Therefore, a method that is consistent with architectural scene recognition is needed to improve the accuracy and efficiency of acquiring semantic information from indoor building 3D models. Summary of the Invention

[0006] In view of the defects in the prior art, the purpose of the present invention is to provide a method and system for acquiring semantic information of a building model.

[0007] According to one aspect of the present invention, a method for acquiring semantic information of a building model is provided, comprising:

[0008] Import the 3D model of the area to be constructed and select the wall to be inspected in the model as the current wall;

[0009] Calculating the geometric outline of the current wall;

[0010] Determine whether there is a porch within the geometric outline of the current wall:

[0011] If it exists, update the geometric outline and the semantic information of the current wall according to the boundary of the porch; if it does not exist, maintain the semantic information of the current wall;

[0012] Calculating the distance information of the current wall to the wall based on the geometric outline of the current wall, and adding the information to the semantic information of the current wall;

[0013] Based on the geometric outline of the current wall, one or more of holes, beams and bay windows of the current wall are identified and added to the semantic information of the current wall.

[0014] Preferably, the calculating the geometric outline of the current wall includes:

[0015] Get the Cartesian coordinates of a point on the wall to be detected, named hitPoint;

[0016] Using the hitPoint as an anchor point, emit rays in the up, down, left, and right directions perpendicular to the wall normal vector to monitor collision events.

[0017] Based on the existence of a ceiling and a floor in the imported three-dimensional model, the coordinates corresponding to the points where the upper and lower longitudinal rays collide are directly converted into the height of the current wall;

[0018] Based on the existence of a positive angle in the imported three-dimensional model, determine whether a collision event occurs between the left and right transverse rays within a preset length range:

[0019] If a collision occurs, the geometric outline of the current wall is obtained based on the coordinates of the collision point and the starting point of the ray;

[0020] If no collision occurs, the preset length is used as the boundary of the current wall to obtain the geometric outline of the current wall.

[0021] Preferably, the determining whether there is a porch within the geometric outline of the current wall: if so, updating the geometric outline and the semantic information of the current wall according to the boundary of the porch; if not, maintaining the semantic information of the current wall; includes:

[0022] Obtain the hitPoint coordinates and the geometric outline of the current wall;

[0023] With the vertical axis where the hitPoint is located as the center, rays are emitted in the direction of the normal vector of the wall in the left and right areas respectively;

[0024] Traverse with a specific step size and monitor whether a collision occurs within the set distance:

[0025] If a collision occurs, the semantic information indicating that the current wall is a porch is added;

[0026] If no collision occurs, the semantic information indicating that the current wall is a facade is supplemented.

[0027] Preferably, the calculating the wall distance information of the current wall includes:

[0028] Receiving the geometric outline of the current wall;

[0029] Starting from the lower left corner of the wall, set traversal starting points in the right and upward directions parallel to the current wall surface as the point matrix for multiple ray emission;

[0030] Traverse all points in the dot matrix, emit rays in the direction of the wall normal vector in turn, monitor collision events, and record collision coordinates;

[0031] A point closest to the current wall from the plurality of collision coordinates is selected, and the distance of the closest point is updated into the semantic information of the current wall as the value of the distance to the wall.

[0032] Preferably, identifying the holes in the current wall and adding them to the semantic information of the current wall includes:

[0033] Receiving the geometric outline of the current wall;

[0034] Starting from the lower left corner of the current wall, set traversal starting points in the right and upward directions parallel to the current wall, and emit rays in the opposite direction of the wall normal vector at each traversal starting point;

[0035] Listen for events where no collision occurred:

[0036] If a collision occurs, it means that no hole is detected and the traversal is continued;

[0037] If no collision occurs, the coordinate point where no collision occurs is used as the coordinate of the lower left corner of the hole;

[0038] Starting from the lower left corner of the hole, set traversal starting points in the right and upward directions parallel to the wall, and emit rays in the opposite direction of the wall normal vector at each starting point;

[0039] Listen for collision events:

[0040] If no collision occurs, it means that the hole boundary has not been detected and the traversal continues;

[0041] If a collision occurs, it means that the hole boundary is detected and added to the semantic information of the current wall.

[0042] Preferably, identifying the bay window on the current wall and adding it to the semantic information of the current wall includes:

[0043] The geometric outline of the receiving hole;

[0044] Take the lower left corner of the hole as the reference point, offset it to the left by a distance a, and set the traversal starting point to the right as the point matrix for emitting rays;

[0045] Listen for events where no collision occurred:

[0046] If a collision occurs, the presence of a floating window is detected and the traversal continues;

[0047] If no collision occurs, the depth information of the floating window is obtained;

[0048] Determine whether the floating window depth information is less than the set process parameters:

[0049] If it is greater, it indicates that there is a bay window at the hole, and the relevant semantics are updated to the current wall.

[0050] Preferably, identifying the beams of the current wall and adding them to the semantic information of the current wall includes:

[0051] If the bay window depth information is less than the set process parameters, it indicates that there is no bay window at the hole, and further determination is made as to whether there is a beam;

[0052] Take the lower left corner of the hole as the reference point, offset it to the left by a distance a, and set the traversal starting point to the right as the point matrix for emitting rays;

[0053] Listen for events where no collision occurred:

[0054] If a collision occurs, the presence of a beam is detected and the traversal continues;

[0055] If no collision occurs, obtain the beam protrusion distance;

[0056] Determine whether the beam protrusion distance is less than the set process parameters:

[0057] If it is greater than, there is a beam at the hole, and the relevant semantics are updated to the wall;

[0058] If it is less than, there is no beam at the hole, it is just a hole.

[0059] Preferably, according to the final semantic information of the current wall, the geometric outline, space size, distance to the wall, and information of holes, bay windows and beams of the current wall are obtained to generate a comprehensive semantic map.

[0060] According to a second aspect of the present invention, there is provided a system for acquiring semantic information of a building model, comprising:

[0061] Wall boundary recognition module: imports a 3D model of the area to be constructed, selects the wall to be detected in the model as the current wall, and calculates the geometric outline of the current wall;

[0062] The porch area detection module determines whether there is a porch within the geometric outline of the current wall. If so, the geometric outline and the semantic information of the current wall are updated according to the boundary of the porch. If not, the semantic information of the current wall is maintained.

[0063] A wall distance acquisition module calculates the wall distance information of the current wall based on the geometric outline of the current wall, and adds the information to the semantic information of the current wall;

[0064] Door and window hole recognition module: Based on the geometric outline of the current wall, identify one or more holes, beams and bay windows on the current wall, and add them to the semantic information of the current wall.

[0065] Preferably, the geometric outline output by the wall boundary recognition module includes boundary coordinate points of the wall; the geometric outline is transmitted to the entrance area detection module and the wall distance acquisition module;

[0066] The porch area detection module outputs the boundary coordinate points and space size of the porch area; and transmits the boundary coordinate points and space size of the porch area to the wall distance acquisition module and the door and window hole recognition module;

[0067] The wall distance acquisition module outputs the distance of the wall directly opposite to the currently identified wall from the current wall; and transmits the distance information to the door and window hole recognition module;

[0068] The door and window hole recognition module outputs the location, size, and type of the holes, and the location and size of the beams and bay windows.

[0069] Compared with the prior art, the embodiments of the present invention have at least one of the following beneficial effects:

[0070] The method and system for acquiring semantic information about building models, as described in embodiments of the present invention, generate a detailed semantic map by comprehensively analyzing the geometric contours of walls, entrances, spatial dimensions of the distance to the wall, and information such as holes, beams, and bay windows through multi-dimensional collaborative operations. This method and system can provide construction robots with rich environmental information to support their planning and execution of complex tasks.

[0071] The semantic information acquisition system for building models in this embodiment of the present invention can import different types of indoor environments as 3D models, demonstrating multi-environment adaptability. Specifically, the entire semantic information acquisition system supports a wide range of indoor environments, including simple and complex architectural structures. The system can flexibly adapt to different scenarios and various interior layouts and architectural styles. It can also update 3D models in real time, perform dynamic processing, and adapt to environments with dynamically changing process types. This enables construction robots to accurately acquire semantic information even in dynamic environments, ensuring real-time and accurate task planning.

[0072] The building model semantic information acquisition system involved in the embodiments of the present invention possesses efficient information processing capabilities. Specifically, by utilizing the raycasting function within the 3D engine and combining it with an efficient collision detection algorithm, the system can rapidly obtain the coordinates and normal directions of collision points. This significantly reduces the amount of computation required, thereby improving the efficiency of information acquisition. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] Other features, objects and advantages of the present invention will become more apparent upon reading the detailed description of non-limiting embodiments with reference to the following drawings:

[0074] Figure 1 is a flow chart of a method for acquiring semantic information of a building model in one embodiment of the present invention;

[0075] Figure 2 This is a flow chart of identifying wall boundaries in a preferred embodiment of the present invention;

[0076] Figure 3 is a schematic diagram of a wall normal vector in a preferred embodiment of the present invention;

[0077] Figure 4 This is a flow chart of detecting the entrance area in a preferred embodiment of the present invention;

[0078] Figure 5 A schematic diagram of detecting the entrance area in a preferred embodiment of the present invention;

[0079] Figure 6 This is a flow chart of obtaining the distance to the wall in a preferred embodiment of the present invention;

[0080] Figure 7 Schematic diagram of the concept of distance to the wall in a preferred embodiment of the present invention;

[0081] Figure 8 A flowchart of identifying a single door or window hole in a preferred embodiment of the present invention;

[0082] Figure 9 This is a flow chart of identifying beams and bay windows in a preferred embodiment of the present invention;

[0083] Figure 10 Schematic diagram of the process of identifying a single door or window hole in a preferred embodiment of the present invention;

[0084] Figure 11 Schematic diagram of the process of identifying a single bay window in a preferred embodiment of the present invention;

[0085] Figure 12 Schematic diagram of the process of identifying a single beam in a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0086] The present invention will be described in detail below with reference to specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those skilled in the art, several variations and improvements can be made without departing from the scope of the present invention. These all fall within the scope of protection of the present invention.

[0087] One embodiment of the present invention provides a method for acquiring semantic information of a building model, such as Figure 1 As shown, the following steps are included:

[0088] Step 1: Import the 3D model of the area to be constructed, select the wall to be inspected in the model as the current wall, and calculate the geometric contour of the current wall.

[0089] Generally, you can select the wall to be detected by hovering the mouse.

[0090] Step 2: Determine whether there is a porch within the geometric outline of the current wall:

[0091] If it exists, update the geometric outline and semantic information of the current wall according to the boundary of the entrance; if it does not exist, maintain the semantic information of the current wall;

[0092] Step 3: Based on the geometric outline of the current wall, calculate the wall distance information of the current wall and add it to the semantic information of the current wall;

[0093] Step 4: Based on the geometric outline of the current wall, identify one or more of the holes, beams, and bay windows on the current wall, and add them to the semantic information of the current wall.

[0094] The above-mentioned embodiment, through multi-dimensional collaborative operations, comprehensively analyzes the geometric contours of the wall, the dimensions of the entrance, the distance to the wall, and information such as holes, beams, and bay windows to generate a detailed semantic map. This method and system can provide construction robots with rich environmental information to support their planning and execution of complex tasks.

[0095] It should be noted that in some embodiments, when it is necessary to identify holes, beams, and bay windows on the current wall, the execution order of step 3 and step 4 can be interchanged, and does not constitute a sole limitation.

[0096] In a preferred embodiment, a preferred process of step 1 is provided, specifically, as follows Figure 2 As shown, to calculate the geometric outline of the current wall, you can use the following steps:

[0097] Step 1.1: Get the Cartesian coordinates of a point on the wall to be detected (current wall) by hovering the mouse, and name it hitPoint.

[0098] Step 1.2, using the hitPoint selected in step 1 as the anchor point, emit collision-monitoring events in four directions perpendicular to the wall normal vector, such as Figure 3 shown.

[0099] Step 1.2 can be implemented through programming. In programming, a monitoring event is a specific action or state change predefined in the program that can be actively monitored and responded to by the code. In 3D software, a collision event is triggered when two or more virtual objects come into contact or overlap in 3D space.

[0100] In step 1.3, based on the existence of a ceiling and a ground in the imported 3D model, the coordinates corresponding to the collision point of the longitudinal ray are directly converted into the height of the wall.

[0101] Step 1.4: Based on the presence of a sun angle in the house design, determine whether a collision event occurs within a specific length range for the horizontal rays:

[0102] If there are no obstacles in the ray's direction that could cause a collision, the ray's return value is set to infinity. However, the concept of infinity doesn't exist in software implementations, so a maximum threshold is pre-set to serve as the specific length range mentioned above. When the ray length exceeds this threshold, collision detection is discontinued, saving time and avoiding invalid detections.

[0103] If a collision occurs, obtain the geometric outline of the wall:

[0104] When two objects collide, the collision event provides the identity and properties of the collision objects, the coordinates of the collision point, and the collision normal. The coordinates of the ray's starting point and the collision point can be used to calculate the wall's outline.

[0105] If no collision occurs, the aforementioned specific length is temporarily used as the current wall boundary to obtain the wall's geometric outline. Walls are typically rectangular, and their outline information includes the coordinates of the top-left, bottom-left, top-right, and bottom-right corners. Based on this, the width and height of the wall can be accurately calculated. These data are considered the geometric shape's outline parameters.

[0106] The external corners mentioned in the above embodiment are architectural features. The corresponding internal corners are recessed corners, such as the angle between a ceiling and surrounding walls. External corners are protruding corners, such as the angle between two walls at a corner in a hallway. The above embodiment ensures the accuracy of wall boundary information by precisely determining the geometric outline of the object to be identified. This provides a solid foundation for subsequent region detection and semantic information supplementation.

[0107] In a preferred embodiment, a preferred process of step 2 is provided, such as Figure 4Specifically, it is determined whether there is a porch within the geometric outline of the wall. If so, the geometric outline and the semantic information of the current wall are updated according to the boundary of the porch. If not, the semantic information of the current wall is maintained, including:

[0108] Step 2.1, obtain the coordinates of the hitPoint point and the geometric outline of the wall to be detected;

[0109] Step 2.2, with the vertical axis where hitPoint is located as the center, emit rays in the direction of the normal vector of the wall in the left and right areas respectively, such as Figure 5 As shown, the red point is the hitPoint, the purple arrow represents the ray group emitted in the direction of the normal vector on the left, and the blue arrow represents the ray group emitted in the direction of the normal vector on the right.

[0110] Step 2.3: Traverse with a specific step size to monitor whether a collision occurs within a specific distance range:

[0111] The specific step size here refers to the spacing between rays, and traversal refers to the sequential emission of all rays and the detection of a collision event. The specific distance range here generally refers to the defined value of the entrance width. Of course, the defined value may vary depending on the scenario.

[0112] If a collision occurs, additional semantic information is detected to indicate that the wall is a porch;

[0113] If no collision occurs, the semantic information that the wall is a facade is additionally detected.

[0114] The collision information obtained in step 2.3 reveals the width, height, and corner coordinates of the entryway's walls. This allows the robot to accurately determine the spatial size of the object to be identified, crop and isolate the entryway area, and assign it new semantic information. This helps construction robots quickly identify and process specific areas in complex environments.

[0115] In a preferred embodiment, a preferred process of step 3 is provided, specifically, as follows Figure 6 and Figure 7 As shown in the figure, the wall distance 1 and the wall distance 2 refer to the depth value of the building wall. Step 3, calculate the wall distance information of the selected wall, and the following steps can be used:

[0116] Step 3.1, receiving the geometric outline of the wall to be detected;

[0117] Step 3.2: Starting from the lower left corner of the wall, set traversal starting points in the right and upward directions parallel to the wall surface as the point matrix for multiple ray emission;

[0118] Step 3.3, traverse all points in the dot matrix, emit rays in the direction of the wall normal vector, monitor collision events, and record the collision coordinates;

[0119] Step 3.4: Filter out the point closest to the detected wall among the multiple collision coordinates, and update the distance of the closest point to the semantic information of the wall as the value of the distance to the wall.

[0120] In a preferred embodiment, a preferred process of step 4 is provided, specifically, as follows Figure 8 and Figure 10 As shown in Figures (a)-(i), the process of identifying door and window holes is as follows:

[0121] Step 4.1, receiving the geometric outline of the detected wall;

[0122] Step 4.2, starting from the lower left corner of the wall, Figure 10 The red point is set as the starting point of the traversal in the right and upward directions parallel to the wall. Figure 10 The blue and red points are the points traversed to the right and upward respectively; at each starting point, a ray is emitted in the opposite direction of the wall normal vector. Figure 10 The purple arrows in the figure represent the emitted rays.

[0123] Step 4.3, listen for events where no collision occurs:

[0124] The non-collision event here is inferred from the collision event. For example, if the propagation speed of the ray is known and the distance to be detected is known, the approximate time required for the ray to collide can be calculated. Therefore, if no collision event is detected within this time period, it is considered a non-collision event.

[0125] If a collision occurs and no holes are detected, return and continue traversal;

[0126] If no collision occurs, the coordinate point where no collision occurs is used as the lower left coordinate of the hole;

[0127] Step 4.4: Starting from the lower left corner of the hole, set traversal starting points in the right and upward directions parallel to the wall, and emit rays from each starting point in the direction opposite to the wall normal vector;

[0128] Step 4.5, listen for collision events:

[0129] If no collision occurs, it means that the hole boundary has not been detected and the traversal continues;

[0130] If a collision occurs, the hole boundary is detected and added to the semantic information of the current wall.

[0131] In some embodiments, it is further determined whether the upper and lower limits of the hole are flush with the wall. If so, the hole is an area that was not detected in the contour recognition stage, the wall contour information is updated, and the hole information is deleted; if not, the hole boundary is further detected and added to the semantic information of the current wall.

[0132] Based on the obtained hole semantic information, we further identify the bay windows and beams, such as Figure 9 and Figure 11 As shown in Figures (a)-(e). Specifically:

[0133] Step 4.5, receiving the geometric outline of the hole;

[0134] Step 4.6, take the lower left corner of the hole as the reference point, Figure 11 The red point in the middle is offset to the left by a distance and the traversal starting point is set to the right. Figure 11 The blue dots in the middle serve as the dot matrix for emitting rays (purple arrows);

[0135] Step 4.7, listen for events where no collision occurs:

[0136] If a collision occurs, the presence of a floating window is detected and the traversal continues;

[0137] If no collision occurs, the depth information of the floating window is obtained ( Figure 11 (purple double arrows in Figure (e));

[0138] Step 4.8: Determine whether the floating window depth information is less than the set process parameters:

[0139] If it is greater, it indicates that there is a bay window at the hole, and the relevant semantics are updated to the current wall.

[0140] If the bay window depth information is less than the set process parameters, it means that there is no bay window at the hole, and further determine whether there is a beam, such as Figure 10 and Figure 12 As shown in Figures (a)-(e). Specifically:

[0141] Step 4.9, take the lower left corner of the hole as the reference point, Figure 12 The red point in the middle is offset to the left by a distance and the traversal starting point is set to the right. Figure 12 The blue dots in the middle serve as the dot matrix for emitting rays (purple arrows);

[0142] Step 4.10, listen for events where no collision occurs:

[0143] If a collision occurs, the presence of a beam is detected and the traversal continues;

[0144] If no collision occurs, obtain the beam protrusion distance;

[0145] Step 4.11, determine the beam protrusion distance ( Figure 12 Is the purple double arrow in the figure (e) less than the set process parameters?

[0146] If it is greater than, there is a beam at the hole, and the relevant semantics are updated to the wall;

[0147] If it is less than, there is no beam at the hole, it is just a hole.

[0148] The above embodiment further supplements the semantic information of the objects to be identified, including filtering, merging, and cropping of holes, as well as the identification of features such as beams and bay windows. This makes the semantic information richer and more detailed, improving the accuracy of task planning.

[0149] The above embodiments use the raycasting function in the 3D engine, combined with an efficient collision detection algorithm, to quickly obtain the coordinates and normal direction of the collision point, which greatly reduces the amount of calculation and improves the efficiency of information acquisition.

[0150] Based on the same inventive concept, another embodiment of the present invention provides a system for acquiring semantic information of a building model, comprising:

[0151] Wall boundary recognition module: Import the 3D model of the area to be constructed, select the wall to be detected in the model by hovering the mouse, use it as the current wall, and calculate the geometric outline of the current wall.

[0152] The wall boundary recognition module ensures the accuracy of wall boundary information by precisely determining the geometric outline of the object to be recognized, which provides a solid foundation for subsequent region detection and semantic information supplementation.

[0153] Porch area detection module: determines whether there is a porch within the geometric outline of the wall; if so, updates the geometric outline and the semantic information of the current wall according to the boundary of the porch; if not, maintains the semantic information of the current wall

[0154] The entrance area detection module accurately identifies the spatial size of the object to be identified, crops and separates the entrance area, and assigns new semantic information. This helps construction robots quickly identify and process specific areas in complex environments.

[0155] Wall distance acquisition module: calculates the wall distance information of the current wall;

[0156] Door and window hole recognition module: Based on distance information, it identifies one or more holes, beams, and bay windows on the current wall and adds them to the semantic information of the current wall.

[0157] The door and window hole recognition module further supplements the semantic information of the objects to be identified, including filtering, merging, and cropping of holes, as well as the recognition of features such as beams and bay windows. This makes the semantic information richer and more detailed, improving the accuracy of task planning.

[0158] The specific implementation techniques of the modules / units in the above examples of the present invention may refer to the corresponding steps of the method for acquiring semantic information of a building model in the above embodiment, and will not be repeated here.

[0159] To further ensure convenient data transmission and storage, a preferred embodiment also employs a storage or transmission module that stores or transmits the semantic information generated by all modules. This module provides reliable data support. This acquired semantic information can be stored and transmitted to the subsequent construction robot's task planning module, ensuring the accuracy and reliability of task planning.

[0160] In some specific embodiments, the above modules are described in detail. Specifically:

[0161] The wall boundary recognition module is used to initially identify the geometric outline of the wall in the Cartesian coordinate system. The data it obtains are the boundary coordinate points of the wall.

[0162] Subsequent data call: pass the boundary coordinate points of the wall to the entrance area detection module and the depth information acquisition module.

[0163] The entryway area detection module identifies the wall's spatial size, crops and separates the entryway area, and updates the wall's boundary coordinates. The data it acquires includes the entryway's boundary coordinates and spatial size.

[0164] Subsequent data call: The boundary coordinate points and space size of the entrance area are passed to the depth information acquisition module and the door and window hole recognition module.

[0165] The wall distance acquisition module calculates the local environment information of the construction area required for path planning. The data it obtains is the distance from the wall directly opposite the currently identified wall to the current wall.

[0166] Subsequent data call: The depth information and local environmental characteristics of the construction area are passed to the door and window hole recognition module and the storage or transmission module.

[0167] The door and window hole recognition module further supplements the semantic information of the wall surface, including filtering, merging, and cropping multiple holes, as well as identifying features such as beams and bay windows. The data it acquires includes the location, size, and type of holes, and the location and size of beams and bay windows.

[0168] Subsequent calls to data: pass details of holes and features to storage or transmission modules.

[0169] Furthermore, in some other embodiments, the data flow between modules is supplemented. The wall boundary recognition module determines and recognizes the geometric outline of the wall, generates the boundary coordinate points of the wall, and transmits them to the entrance area detection module and the wall distance acquisition module.

[0170] The entrance area detection module identifies the spatial size of the wall based on the boundary coordinate points of the wall, crops and separates the entrance area, generates the boundary coordinate points and spatial size of the entrance area, and passes them to the wall distance acquisition module and the door and window hole recognition module.

[0171] The wall distance acquisition module calculates the depth information and local environmental characteristics of the construction area based on the boundary coordinate points of the wall and the boundary coordinate points of the entrance area, and transmits them to the door and window hole recognition module and the storage or transmission module.

[0172] The door and window hole recognition module further supplements the semantic information of the wall based on depth information and local environmental characteristics, including the filtering, merging, and cropping of holes, as well as the recognition of features such as beams and bay windows. It generates detailed information on holes and features and passes it to the storage or transmission module.

[0173] The storage or transmission module stores or transmits the semantic information generated by all modules to the task planning module of the construction robot.

[0174] A construction model is usually a house type composed of multiple rooms, including but not limited to the master bedroom, second bedroom, study, kitchen, bathroom, and other room areas. The spatial coordinate information of the wall is used to determine which walls belong to the same room. After forming a room, the room size information is used to infer the type of the above-mentioned room. In order to verify the feasibility and effectiveness of the semantic information acquisition method and system of the building model in the above embodiment, in a specific embodiment, a specific wall is identified, and the process is as follows:

[0175] S1: Call the recognition interface and import the model to be recognized. The model is a bedroom of a certain residential unit with unknown dimensions.

[0176] S2: Move the mouse and hover over the wall to be identified, and obtain the coordinates of the anchor point hitPoint (0, y h ,z h ), and the wall normal vector (1,0,0).

[0177] S3: (0,y h ,z h ) as the starting point, and emit rays in four directions: (0,1,0), (0,-1,0), (0,0,1), and (0,0,-1).

[0178] S4: The coordinates of the collision position of the ray in the (0,1,0) direction are (0,3,z h ), the coordinates of the collision position of the ray in the (0,-1,0) direction are (0,-1.2,z h ), the coordinates of the collision position of the ray in the (0,0,1) direction are (0,y h ,2.8), the coordinates of the collision position of the ray in the (0,0,-1) direction are (0,y h ,0).

[0179] S5: The rough geometric information of the wall is initially obtained from the four collision points. The coordinates of the four vertices are (0, -1.2, 0), (0, 3, 0), (0, -1.2, 2.8), and (0, 3, 2.8).

[0180] S6:(0,y h ,z h ) as the starting point, ray is emitted towards the wall normal vector (1,0,0), and no collision event within a distance of 2 meters is detected.

[0181] S7: (0,y h ,0) as the starting point, several anchor points have been set up in an incremental manner for the y and z values, and a ray is emitted to (1,0,0). After traversal, no collision event within a distance of 2 meters is detected.

[0182] S8: Starting from (0,-1.2,0), set traversal points in the directions of (0,1,0) and (0,0,1), with upper limits of 3.2 meters and 2.8 meters respectively.

[0183] S9: Starting from the traversal point generated in step S8, a ray is emitted toward the normal vector (1, 0, 0) of the wall, collision events are monitored, and the distance when the collision occurs is recorded.

[0184] S10: Filter out the minimum distance of 3 meters from a number of distances.

[0185] S11: Starting from (0.2, -1.2, 0), set traversal points in the directions of (0, 1, 0) and (0, 0, 1), with upper limits of 3.2 meters and 2.8 meters respectively.

[0186] S12: Starting from the traversal point generated in step S11, a ray is emitted toward (-1, 0, 0), and no collision event is detected.

[0187] S13: After traversing all points, no points without collision are detected, so there are no door or window holes on this wall.

[0188] S14: Outputting semantic information of the wall. The semantic information of the wall on which the mouse is hovering is as follows: left corner coordinates (0, -1.2, 0), wall width 3.2 meters, wall height 2.8 meters, inner corners on both sides, distance to the opposite wall 3 meters, and no door or window holes on the wall.

[0189] It should be noted that the various modules in the semantic information acquisition system of the building model of the construction robot provided in the above embodiments of the present invention correspond to the steps of the semantic information acquisition method of the building model in any of the above embodiments. Those skilled in the art can refer to the technical features of the steps of the semantic information acquisition system method of the building model to implement the corresponding modules in the semantic information acquisition system of the building model, which will not be repeated here.

[0190] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0191] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0192] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0193] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0194] The above describes specific embodiments of the present invention. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art may make various modifications or variations within the scope of the claims without affecting the essence of the present invention. The above preferred features may be used in any combination as long as they do not conflict with each other.

Claims

1. A method for acquiring semantic information of a building model, characterized in that: include: Importing a three-dimensional model of the area to be constructed, selecting a wall surface to be inspected in the three-dimensional model as the current wall surface, and calculating the geometric contour of the current wall surface; Determine whether there is a porch within the geometric outline of the current wall: If it exists, update the geometric outline and the semantic information of the current wall according to the boundary of the porch; if it does not exist, maintain the semantic information of the current wall; Calculating the distance information of the current wall to the wall based on the geometric outline of the current wall, and adding the information to the semantic information of the current wall; Based on the geometric outline of the current wall, one or more of holes, beams and bay windows of the current wall are identified and added to the semantic information of the current wall.

2. The method for acquiring semantic information of a building model according to claim 1, characterized in that: The calculating the geometric outline of the current wall includes: Get the Cartesian coordinates of a point on the wall to be detected, named hitPoint; Using the hitPoint as an anchor point, emit rays in the up, down, left, and right directions perpendicular to the wall normal vector to monitor collision events. Based on the existence of a ceiling and a floor in the imported three-dimensional model, the coordinates corresponding to the points where the upper and lower longitudinal rays collide are directly converted into the height of the current wall; Based on the existence of a positive angle in the imported three-dimensional model, determine whether a collision event occurs between the left and right transverse rays within a preset length range: If a collision occurs, the geometric outline of the current wall is obtained based on the coordinates of the collision point and the starting point of the ray; If no collision occurs, the preset length is used as the boundary of the current wall to obtain the geometric outline of the current wall.

3. The method for acquiring semantic information of a building model according to claim 2, characterized in that: Determining whether there is a porch within the geometric outline of the current wall: if so, updating the geometric outline and semantic information of the current wall according to the boundary of the porch; If not, maintain the semantic information of the current wall; including: Obtain the hitPoint coordinates and the geometric outline of the current wall; With the vertical axis where the hitPoint is located as the center, rays are emitted in the direction of the normal vector of the wall in the left and right areas respectively; Traverse with a specific step size and monitor whether a collision occurs within the set distance: If a collision occurs, the semantic information indicating that the current wall is a porch is added; If no collision occurs, the semantic information indicating that the current wall is a facade is supplemented.

4. The method for acquiring semantic information of a building model according to claim 1, wherein: The calculating the wall distance information of the current wall includes: Receiving the geometric outline of the current wall; Starting from the lower left corner of the wall, set traversal starting points in the right and upward directions parallel to the current wall surface as the point matrix for multiple ray emission; Traverse all points in the dot matrix, emit rays in the direction of the wall normal vector in turn, monitor collision events, and record collision coordinates; A point closest to the current wall from the plurality of collision coordinates is selected, and the distance of the closest point is updated into the semantic information of the current wall as the value of the distance to the wall.

5. The method for acquiring semantic information of a building model according to claim 1, wherein: Identifying holes in the current wall and adding them to the semantic information of the current wall includes: Receiving the geometric outline of the current wall; Starting from the lower left corner of the current wall, set traversal starting points in the right and upward directions parallel to the current wall, and emit rays in the opposite direction of the wall normal vector at each traversal starting point; Listen for events where no collision occurred: If a collision occurs, it means that no hole is detected and the traversal is continued; If no collision occurs, the coordinate point where no collision occurs is used as the coordinate of the lower left corner of the hole; Starting from the lower left corner of the hole, set traversal starting points in the right and upward directions parallel to the wall, and emit rays in the opposite direction of the wall normal vector at each starting point; Listen for collision events: If no collision occurs, it means that the hole boundary has not been detected and the traversal continues; If a collision occurs, it means that the hole boundary is detected and added to the semantic information of the current wall.

6. The method for acquiring semantic information of a building model according to claim 5, characterized in that: Identifying the bay window on the current wall and adding it to the semantic information of the current wall includes: The geometric outline of the receiving hole; Take the lower left corner of the hole as the reference point, offset it to the left by a distance a, and set the traversal starting point to the right as the point matrix for emitting rays; Listen for events where no collision occurred: If a collision occurs, the presence of a floating window is detected and the traversal continues; If no collision occurs, the depth information of the floating window is obtained; Determine whether the floating window depth information is less than the set process parameters: If it is greater, it indicates that there is a bay window at the hole, and the relevant semantics are updated to the current wall.

7. The method for acquiring semantic information of a building model according to claim 6, characterized in that: Identifying the beams of the current wall and adding them to the semantic information of the current wall includes: If the bay window depth information is less than the set process parameters, it indicates that there is no bay window at the hole, and further determination is made as to whether there is a beam; Take the lower left corner of the hole as the reference point, offset it to the left by a distance a, and set the traversal starting point to the right as the point matrix for emitting rays; Listen for events where no collision occurred: If a collision occurs, the presence of a beam is detected and the traversal continues; If no collision occurs, obtain the beam protrusion distance; Determine whether the beam protrusion distance is less than the set process parameters: If it is greater than, there is a beam at the hole, and the relevant semantics are updated to the wall; If it is less than, there is no beam at the hole, it is just a hole.

8. The method for acquiring semantic information of a building model according to claim 1, wherein: According to the final semantic information of the current wall, the geometric outline, space size, distance to the wall, and information of holes, bay windows and beams of the current wall are obtained to generate a comprehensive semantic map.

9. A semantic information acquisition system for a building model, characterized in that: include: Wall boundary recognition module: import the 3D model of the area to be constructed and select the wall to be detected in the model as the current wall; Calculating the geometric outline of the current wall; The porch area detection module determines whether there is a porch within the geometric outline of the current wall. If so, the geometric outline and the semantic information of the current wall are updated according to the boundary of the porch. If not, the semantic information of the current wall is maintained. A wall distance acquisition module calculates the wall distance information of the current wall based on the geometric outline of the current wall, and adds the information to the semantic information of the current wall; Door and window hole recognition module: Based on the geometric outline of the current wall, identify one or more holes, beams and bay windows on the current wall, and add them to the semantic information of the current wall.

10. The semantic information acquisition system of a building model according to claim 9, characterized in that: The geometric outline output by the wall boundary recognition module includes the boundary coordinate points of the wall; the geometric outline is transmitted to the entrance area detection module and the wall distance acquisition module; The porch area detection module outputs the boundary coordinate points and space size of the porch area; and transmits the boundary coordinate points and space size of the porch area to the wall distance acquisition module and the door and window hole recognition module; The wall distance acquisition module outputs the distance of the wall directly opposite to the currently identified wall from the current wall; and transmits the distance information to the door and window hole recognition module; The door and window hole recognition module outputs the location, size, and type of the holes, and the location and size of the beams and bay windows.

Images