CN111179412B - Automatic processing method and system for house type graph - Google Patents
Automatic processing method and system for house type graph Download PDFInfo
- Publication number
- CN111179412B CN111179412B CN201911273378.8A CN201911273378A CN111179412B CN 111179412 B CN111179412 B CN 111179412B CN 201911273378 A CN201911273378 A CN 201911273378A CN 111179412 B CN111179412 B CN 111179412B
- Authority
- CN
- China
- Prior art keywords
- room
- layer
- rooms
- polygon
- house type
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Abstract
The invention relates to a method and a system for automatically processing a house type graph, belonging to the technical field of computer aided design. The method for automatically generating the house type space region is provided, a designer can automatically generate the region by drawing a wall body, the bottom point of the wall body is the same as the edge point of the region, the error of numerical precision cannot occur, and the method is favorable for expanding and using other functionalities of home decoration design software; meanwhile, the invention also provides an automatic space grouping method suitable for villas or small high-rise dwelling houses, which can save resource investment, help designers to design schemes like drawing common flat-storey houses, automatically calculate the grouping relation of floor spaces and reduce the working pressure of the designers; the invention also provides a method for quickly positioning the local maximum rectangle and the region center position of the house type space based on deep learning.
Description
Technical Field
The invention relates to a method and a system for automatically processing a house type graph, belonging to the technical field of computer aided design.
Background
The existing home decoration design software has the disadvantages that the generation effect of a space region is not ideal: some spaces surrounded by thin walls are also marked as regions (actually, the spaces are inner spaces of flues or other components, as shown in fig. 1, in a room, since some common accessories also need to be compiled into a thin polygon, and many software can mistakenly classify such a thin polygon region into a room when identifying the room region of a user-type diagram), and some software manually draws a polygon generation region through a designer, which can cause numerical errors between the region edge and the wall edge and influence the use of other functions (for example, a commodity under the region space is judged, when the commodity is a wall painting hanging on a wall, and the projection of the central point of the wall painting model on the ground may not fall inside the region polygon manually drawn by the designer).
The existing home decoration design software is usually suitable for ordinary houses with simple flat floors, however, the home decoration design software is not well compatible for villas or buildings with small high-rise residential types, and meanwhile, for designers, drawing high-rise buildings usually requires higher professional skills and more time investment cost than drawing flat floors.
In the next design step, the existing home decoration software has no function of locating the spatial local maximum rectangle (the maximum inscribed rectangle of a room is usually an important layout position in the room, and can be used for placing specific furniture, setting a navigation point or designing an automatically displayed camera position, for example, the maximum local rectangle is taken as the base and a camera observation point is set in the patent CN 109960850A), but many businesses of the home decoration software are often closely related to the spatial region, especially the local maximum rectangle, where the main furniture can be placed, and the like. The same central position of the area is also an important element of home decoration design software, for example, a user type space application label is displayed on a two-dimensional layer, and a reasonable camera is placed when a scheme is rendered on a three-dimensional layer, which are closely related to the central position of the area. The method usually adopts a general geometric figure method for calculating the local maximum rectangle and the approximate center position in the house type space, and the method usually calculates through constraint conditions of geometric regularity and the like, so that special house types are inevitably met, rules are not covered, the local maximum rectangle is inaccurately positioned, the optimal approximate center position is calculated to be deviated, and the like.
Disclosure of Invention
The first technical problem to be solved by the invention is as follows: a designer can automatically generate the area by drawing the wall, the bottom point of the wall is the same as the edge point of the area, the error of numerical precision cannot occur, and the method and the system are favorable for expanding and using other functionalities of home decoration design software.
The second technical problem to be solved by the invention is: the method and the system for automatically grouping the spaces, which are suitable for villas or small high-rise dwelling houses, can save resource investment, help designers to design schemes like drawing common flat-storey houses, automatically calculate the grouping relation of floor spaces and reduce the working pressure of the designers.
The third technical problem to be solved by the invention is: a method and a system for rapidly positioning the local maximum rectangle and the region center position of a house type space based on deep learning are provided.
The technical scheme is as follows:
an automatic house type graph processing method, comprising a step for generating a room in a house type graph, comprising:
s101, drawing a multi-wall body;
s102, obtaining a central line of a slender rectangle of the wall;
s103, taking any central line, finding the central lines which can be mutually closed into a closed area, and grouping the central lines into a group; sequentially traversing and searching other central lines, grouping the central lines which can be mutually closed into a closed area, and grouping the central lines which cannot be closed with other central lines into an independent group;
and S104, mutually closing the inner walls of the walls corresponding to all the central lines which can be closed into one area into a closed area to serve as a room area.
In one embodiment, the midline refers to the longer midline of the interior of an elongated rectangle when the wall is in the shape of the rectangle.
In one embodiment, further comprising: and deleting the areas with the area smaller than a certain threshold value from the floor plan.
In one embodiment, the certain threshold is 0.25 square meter in area.
In one embodiment, the step 103 of searching for the central lines mutually closed as a closed area includes: and judging whether the end points of the two middle lines are smaller than a threshold value or not, and connecting the head and tail end points of the middle lines in one closed area to be smaller than the threshold value in sequence.
In one embodiment, further comprising: the method for classifying the rooms in the multi-layer house type graph comprises the following steps:
s201, acquiring room data in a house type graph, wherein the data comprises the shape and the position of a closed area of each room, the function of the room and the ID number of the room;
s202, selecting one room as an initial room, searching rooms adjacent to the initial room, classifying the rooms into a set, and traversing and searching the found rooms sequentially to obtain the adjacent rooms, classifying the rooms into the set until the number of the rooms in the set is not changed;
s203, circularly executing the step S202 for the rooms which are not in the set again, and classifying according to the adjacent relation until all the rooms in the indoor graph are classified;
and S204, designing rooms in the set classified into one type according to a plane floor plan mode.
In one embodiment, S204 further includes: the floor on which the room in each category set is located is determined.
In one embodiment, the method for determining an adjacent room in step S202 includes:
selecting all contour lines of a room;
making a vertical extension line segment with a certain distance from the midpoint on the contour line to the outside of the room;
two rooms are considered to be adjacent if the end points of the vertical extension line fall inside the other room.
In one embodiment, further comprising: the method for automatically positioning the maximum local rectangle in the user-type graph comprises the following steps:
s301, obtaining ordered vertex coordinates of the room polygon, and calculating four vertex coordinates of the maximum local rectangle of the room polygon to serve as training sample data;
s302, training a neural network model by adopting training sample data by taking the ordered vertex coordinates of the room polygon obtained in S301 as input values of a neural network and the four vertex coordinates of the maximum local rectangle as output values;
and S303, inputting the ordered vertex coordinates of the room polygon to be processed into the trained neural network model in the S302 to obtain the four vertex coordinates of the maximum local rectangle.
In one embodiment, in S301, the ordered vertex coordinates of the room polygon need to be translated so that the plane coordinate value of one vertex is the origin, and after the coordinates of the four vertices of the largest local rectangle are obtained in S302, the coordinates need to be translated in the opposite direction and at the same distance as the ordered vertex coordinates of the room polygon.
In one embodiment, the neural network model is a BP neural network.
In one embodiment, further comprising: a method for automatically locating an approximate center point of a room in a floor plan, comprising the steps of:
s401, obtaining ordered vertex coordinates of a room polygon, and calculating approximate center point coordinates of the room polygon to serve as training sample data;
s402, training a neural network model by adopting training sample data by taking the ordered vertex coordinates of the room polygon obtained in S401 as input values of a neural network and the approximate central point coordinates as output values;
and S403, inputting the ordered vertex coordinates of the room polygon to be processed into the trained neural network model in S302 to obtain approximate central point coordinates.
In one embodiment, in S401, the ordered vertex coordinates of the room polygon need to be translated so that the plane coordinate value of one vertex is the origin, and after the approximate center point coordinate is obtained in S402, the translation direction is opposite to the translation direction of the ordered vertex coordinates of the room polygon and the distance is the same, and the translation process needs to be performed in reverse.
In one embodiment, the neural network model comprises:
a first layer for receiving the resulting ordered vertex coordinate data of the room polygon;
the second layer is connected with each layer and used for processing the numerical value output by the first layer, and the number of unit nodes of the second layer is the same as the number of vertexes of the room polygon;
the hidden layer is connected with the second layer and used for processing the numerical value output by the second layer;
the output layer is connected with the hidden layer and used for processing the data of the hidden layer and outputting a result;
in one embodiment, the neural network model is a BP neural network.
Advantageous effects
(1) The invention can reduce the cost of using and learning by the user in the house decoration design process, can lead the designer to draw the house type like a common drawing scheme, help the designer to automatically group the spaces of villas or small high-rise residences into a group, and automatically record the adjacent relation between the spaces, can quickly position specific floors and specific certain space information, and improve the working efficiency of the designer.
(2) The method can enable a designer to automatically generate the area by drawing the wall in the process of home decoration, does not need the designer to manually draw the area, improves the working efficiency of the designer, has the same bottom point and edge point of the area, does not have numerical precision errors, and is beneficial to expanding and using other functionalities of home decoration design software (such as judging the attribution of a model in the area, judging the attribution area of a pendant on the wall and the like).
(3) The invention can enable a user to efficiently and accurately calculate the optimal approximate center point of the house type space region and the coordinates of four ordered point of the local maximum rectangle in the house decoration design process. Due to the training and calculation of large sample data, the compatibility of calculation results of the house type space regions with different geometric characteristics is higher than that of conventional regular characteristic calculation. Therefore, the method can better serve the services related to the spatial layout by using the local maximum rectangle in the home decoration design software and the services related to the optimal approximate central point of the positioning area space.
Drawings
FIG. 1 is a house layout under a special condition
FIG. 2 is a flow chart of the generation of a residential space region
FIG. 3. flow of line grouping in wall
FIG. 4 is a method of matching wall centerline end points to wall centerline build areas
FIG. 5 is a schematic view of a wall body (BE is a wall center line, AF and CD are two side edge points of the wall body)
Figure 6. midline closed area schematic.
FIG. 7 is a flow chart for automatic spatial grouping based on villa or small high-rise dwelling size data
FIG. 8 is a schematic view of a region and its accompanying wall and door opening
FIG. 9 is a schematic diagram of the area where the extension point of the two-dimensional center point of the wall is determined
FIG. 10 is a schematic view of the upper and lower layers of a villa
FIG. 11 is a flow chart of pre-processing of training data for calculating the optimal approximate center point of a region
FIG. 12 is a diagram of pre-processing of local maximum rectangle training data
FIG. 13 is a diagram of a multi-layer neural network model structure for calculating the optimal approximate center point of a region
FIG. 14 is a multi-layer neural network model structure for calculating local maximum rectangles of regions
FIG. 15 is a schematic diagram of the optimal approximate center point of the area and the local maximum rectangle of the area ABCD is the maximum rectangle of the bedroom area, P is the schematic diagram of the optimal approximate center point of the area
Detailed Description
The following describes a method for automatically generating a house space area:
the flow chart of the generation of the dwelling space area is shown in figure 1; the flow of the line grouping in the wall is shown in fig. 2; the method of matching the end points of the wall centerline to the build area of the wall centerline is shown in fig. 3.
I. The generation process of the house type space region comprises the following steps:
1. drawing a plurality of sections of wall bodies;
2. acquiring all wall center lines drawn in the steps, wherein the wall center line refers to a longer middle line in a long and thin rectangle when the wall is used as the long and thin rectangle; through the step, the obtained central lines of all the wall bodies can be obtained; various representations of wall midline may be as shown in fig. 5.
3. Performing closed detection and closed grouping creation on the wall center line in the above steps, wherein a specific closed detection grouping method is shown in fig. 3; the main process comprises the following steps: taking a central line, traversing from other central lines, and finding the central line which can be mutually connected end to form a closed area with the central line, thereby meaning that the wall body to which the central lines belong can be closed into a room; as shown in fig. 6, the central lines of the four-sided solid-line quadrilateral walls may be mutually closed to form a closed polygon, so that the four central-line closed regions may be defined as a group. Then, sequentially traversing and searching other central lines, and grouping the central lines which can be closed into a closed area into a group; if during the search a midline is found that cannot be closed with other midlines, it is defined as a separate set of midlines (usually these are separate walls).
4. Matching the center line end points of the wall body with the inner lines of the wall, wherein the specific matching method is shown in figure 4; the method mainly comprises the following steps: for any midline in a group, find the wall of the total wall body close enough to it, and find the wall of the group close to the total midline, and close the inner lines of the wall bodies to each other to form an area, which is used as a closed room. The room is determined by this method.
5. Sequentially placing wall internal lines matched with the wall central line into a constructed closed area;
6. closed regions that are not properly filtered, i.e., regions having an area of 0.25 square meters or less, are removed and marked as non-regions. Through the steps, a closed room is obtained by using the center line of the wall as the basis, and the problem that the small polygon is also judged as the room does not occur.
The method of centerline grouping in the wall is as follows:
1. obtaining the central lines of all walls;
2. randomly selecting the center line of the wall;
3. and searching all wall center lines which can form a closed loop with the wall center line from the rest wall center lines, wherein the steps are as follows:
3.1. selecting a certain end point of the center line of the wall in the step 2 (we take the end point as an example for explanation);
3.2. traversing the remaining wall central lines, and calculating the Distance between the end points (the starting point and the end point) of the wall central line and the end points in the step 3.1;
3.3. if the Distance in the above step is less than or equal to a certain minimum Distance constraint value (assumed to be set to 0.1mm) [ proximity between the main quantitative end points ], connecting the wall centerline with the wall centerline in step 3.1 in sequence;
3.4. similar steps 3.2 to 3.3 as described above are continued until all wall centerlines have been processed,
if the central lines of a plurality of walls are connected end to end in sequence to form a closed loop, the central lines are grouped into the same group;
if the closed loop can not be formed end to end, the closed loop is classified into an independent group.
4. And if all the wall central lines are completely grouped, completing grouping. If there are no wall center lines grouped, the flow returns to the above process, and step 2 to step 3 are executed in a loop until all wall center lines are grouped.
The following describes an automatic spatial grouping method applicable to villas or small high-rise dwelling types:
in general, home type data imported by home decoration software can only reflect some plane home type data; when a design drawing of a villa or a small high-rise building is introduced, a multi-story building appears on the same plane, so that a designer cannot design a complete house structure as usual.
In the following method, the data of the villa or the small high-rise house type imported in the house decoration design software is processed into a single plane house type design drawing which is good for designers.
I. The process of automatic spatial grouping based on villa or small high-rise dwelling size data is shown in fig. 7:
1. the loading scheme acquires data of the whole house type area, including wall data, ground area contour point set data, space identification RoomID, space application RoomUage and the like;
2. acquiring information of any space (FirstArea) in the whole scheme, including a contour point set of a ground area, and information of a door opening and a window opening;
3. storing the firstreata space in the step 2 into a calculated group (calibrated group);
4. calculating the information of all the spaces adjacent to the space FirstArea, including the space identification of the space adjacent to the space, the ground area point set of the corresponding space and the corresponding adjacency relation, and grouping the spaces into the same Group (Group1), wherein the specific method for judging the spatial adjacency relation is described in the section II; the purpose of the step can be realized by taking FirstArea as a starting room and traversing to other adjacent rooms;
5. sequentially selecting a space OneRoom (except first tarea) from the Group1 in the step 4, calculating all adjacent space information of the space, and simultaneously storing the OneRoom into a calculated Group (calibrated Group); the purpose of this step can be to further go around, starting from the respective rooms adjacent to FirstArea, finding other rooms that continue to be adjacent to them;
6. judging whether the space number (recorded as Num1) of the calculated group (calculated group) in the flow is equal to the number (recorded as Num2) of all the spaces of the house type, and if so, ending the calculation; if equal, the house type is a single-layer house type, the rooms are adjacent in sequence, and after the rooms are searched in an outward traversing manner, the obtained total number of the rooms is necessarily equal to the total number of the rooms of the house type, so that the calculation can be finished.
If not, the following steps are continuously executed.
7. And calculating the adjacent space information of the sub-elements of the Group1 Group, including the ground area point set, the space identification of the adjacent space of the sub-elements, the space usage of the adjacent space of the sub-elements, the associated wall data of the adjacent space of the sub-elements and the like. Storing the space identification of the subspace into a calculated group (calledgroup), and iteratively updating the calledgroup;
8. sequentially calculating the adjacent space information of the subspace in the step 7, and iteratively updating the calculated group (calledgroup) until the number of the calledgroup is not changed any more; in the step, the above adjacent rooms can be continuously searched outwards to find the adjacent room, and when the number of all the rooms is not changed any more, the room represents that the search is finished and no other adjacent rooms exist;
9. marking the callated group with the number which is not increased in the process as a first group space set (marked as Firstgroup);
10. it is determined whether the number of firstgroups in the above step is the same as the number of the entire subscriber space,
10.1. if the two are the same, stopping calculation; the purpose of this step is to judge whether the number in the set of rooms that have been determined to be adjacent to each other is equal to the number of all rooms in the house type, if equal, it represents a single-layer design, and can be classified as an integral house type; if not, it represents that some other rooms exist, possibly classified as another layer, so the following search process continues:
10.2. if the difference is not the same, clearing the calculated group (calllatedgroup), executing the flow similar to the step 2 to the step 9, randomly selecting a space from the spaces of the calculated group, calculating the information of the adjacent space, successively analogizing the adjacent spaces of the subspaces of the executed subspaces to obtain a new calculated group (calllatedgroup), judging whether the sum of the number of the first group FirstGroup and the number of the new calculated group is equal to the number of the spaces of the whole house type, and if the sum is equal, stopping the calculation; if not, the similar process is continuously executed until all the spaces are grouped. The purpose of this step, can carry on the similar search to other rooms sequentially, and continue classifying these unclassified rooms according to the adjacent principle, when getting a adjacent classification set, represent the room set in another floor probably;
11. and reasonably inducing and integrating the grouped space information, recording the space of each group and the adjacency relation among the spaces, and recording the information such as the wall height corresponding to the space elements of each group. The hierarchical relation of the floors is sorted from small to large through the wall height, the smallest floor corresponds to the first floor of the villa and is sequentially increased, and the largest floor corresponds to the highest floor of the villa. Fig. 10 is a plan design effect drawing of a villa drawn by a designer, where the left side is a first floor and the right side is a second floor. Through the steps, the rooms with adjacent relations are sequentially searched and classified in a traversal mode, and each obtained class represents a layer of rooms. So far, the designer can design the house type according to the conventional single house type structure.
Method of determining the neighbourhood of an area (fig. 9 is a schematic diagram of the area where the extension of the two-dimensional centre point of the wall falls):
1. acquiring all the attached walls of the area (fig. 8 is a schematic diagram of the area and the attached walls and door opening);
2. traversing each of the walls:
2.1. calculating a two-dimensional central point of the wall;
2.2. calculating the direction vector of the central axis of the wall body, and carrying out vector normalization processing;
2.3. rotating the vector by 90 degrees anticlockwise or clockwise to obtain a rotated unit vector;
2.4. extending the two-dimensional center point along the rotated unit vector for a distance (usually set to be 30-40 cm) to obtain an extension point P1, and similarly extending along the opposite direction of the rotation vector for the same distance to obtain another opposite extension point P2;
2.5. and judging whether the points P1 and P2 in the steps are in a certain region (except for the region of the point P1 and the point P2 in the steps), if so, determining that the region in which the extension point falls is the adjacent region of the region, and keeping the region which is stored in the adjacent region set from continuing to store.
Following method for quickly positioning local maximum rectangle and region center position of house type space based on deep learning
The description is as follows:
I. calculating the preprocessing flow of the training data of the optimal approximate center point of the region:
1. reading data of a region polygon, namely a plurality of vertex coordinates;
2. reading the coordinates of the approximate center point of the area marked manually and recording and storing the coordinates;
3. translating the polygon in the step 1 to enable a first point of the polygon to be located at an origin, namely subtracting the coordinate of the first point from all vertex coordinates of the polygon in numerical value to finish translation;
4. subtracting the first point of the polygon in the step 3 from the approximate center point of the area manually marked in the step 2 to finish the translation processing of the manually marked point;
5. using the polygon data after translation in the process as input of a training sample and using the approximate center point data after translation as gold standard output of the training sample;
II, preprocessing the training data of the local maximum rectangle of the region:
1. reading data of a region polygon, namely a plurality of vertex coordinates;
2. reading four sequence mark points of a local maximum rectangle of the manually marked area, and recording and storing the four sequence mark points;
3. translating the polygon in the step 1 to enable the first point of the polygon to be located at the origin, namely subtracting the coordinate of the first point from all the vertex coordinates of the polygon in numerical value to finish translation;
4. calculating the coordinates of the four translated sequence points of the manually marked local maximum rectangle, namely, subtracting the first vertex coordinate of the polygon similarly to the step 3;
5. taking the translated polygon data, namely coordinates of a plurality of vertexes as input, and taking coordinates of four sequential points of the translated local maximum rectangle as output of a training sample golden standard;
calculating the structure of the multilayer neural network model of the optimal approximate central point of the region, and specifically explaining the structure as follows:
1. the first layer of the network structure is the input layer, i.e. the coordinates of the vertices of the preprocessed polygons, i.e. a vertex array
2. The number of unit nodes of a second layer of the network structure is equal to the number of vertexes of a polygon in the first layer, and the input array data in the first layer enters the second layer to be continuously processed after being subjected to function processing; in practical situation, adjusting the number of unit nodes of the second layer according to the number of the polygon vertices;
3. the third layer to the sixth layer of the network structure are used as middle layers, and the nodes are connected in a full connection mode;
4. the last layer of the network structure is the output layer, which corresponds to the best approximate center point of the preprocessed polygon.
The specific embodiment of the neural network model is not particularly limited, and examples thereof include: a back propagation neural network model of BP type can be employed.
Calculating the structure of the multilayer neural network model of the local maximum rectangle of the region, and specifically explaining the structure as follows:
1. the first layer of the network structure is the input layer, i.e. the coordinates of the vertices of the preprocessed polygons, i.e. a vertex array
2. The number of unit nodes of a second layer of the network structure is equal to the number of vertexes of a polygon in the first layer, and the input array data in the first layer enters the second layer to be continuously processed after being subjected to function processing; in practical situation, adjusting the number of unit nodes of the second layer according to the number of the polygon vertices;
3. the third layer to the sixth layer of the network structure are used as middle layers, and the nodes are connected in a full connection mode;
4. the last layer of the network structure is an output layer, and four point coordinates of the ordered local maximum rectangle of the preprocessed polygon correspond to the output layer.
The specific embodiment of the neural network model is not particularly limited, and examples thereof include: a back propagation neural network model of BP type can be employed.
V, calculating the optimal approximate center point of the area:
1. after training through large sample house type space data, obtaining a trained network model;
2. on the basis of the step 1, newly providing a house type area, and preprocessing the vertex data of the area, namely subtracting a first vertex coordinate;
3. the points processed in the step 2 are plugged into the network model trained in the step 1 to calculate to obtain the coordinates of an output point;
4. and (4) translating the point coordinates in the step (3), namely adding the coordinates of the first point of the initial polygon to obtain a final point, wherein the final point is used as the coordinates of the optimal approximate center point of the area.
Post-processing of calculating local maximum rectangles of the regions:
1. after training through large sample house type space data, obtaining a trained network model;
2. on the basis of the step 1, newly providing a house type area, and preprocessing the vertex data of the area, namely subtracting a first vertex coordinate;
3. the points processed in the step 2 are plugged into the network model trained in the step 1 to calculate coordinates of four output points;
4. and (3) translating the point coordinates in the step (3), namely adding the coordinates of the first point of the initial polygon to obtain the coordinates of the last four points, and taking the point set as the coordinates of the four points of the local maximum rectangle.
Claims (8)
1. An automatic processing method for house type graph, which is characterized in that the method comprises the steps of generating rooms in the house type graph, and comprises the following steps:
s101, drawing a multi-wall body;
s102, obtaining a central line of a slender rectangle of the wall;
s103, taking any central line, finding the central lines which can be mutually closed into a closed area, and grouping the central lines into a group; sequentially traversing and searching other central lines, grouping the central lines which can be mutually closed into a closed area, and grouping the central lines which cannot be closed with other central lines into an independent group;
s104, mutually closing the inner walls of the walls corresponding to all the central lines which can be closed into one area into a closed area as a room area;
further comprising: the method for classifying the rooms in the multi-layer house type graph comprises the following steps:
s201, acquiring room data in a house type graph, wherein the data comprises the shape and the position of a closed area of each room, the function of the room and the ID number of the room;
s202, selecting one room as an initial room, searching rooms adjacent to the initial room, classifying the rooms into a set, and traversing and searching the found rooms sequentially to obtain the adjacent rooms, classifying the rooms into the set until the number of the rooms in the set is not changed;
s203, circularly executing the step S202 again for the rooms which are not classified in the set, and classifying according to the adjacent relation until all the rooms in the floor plan are classified completely;
and S204, designing rooms in the set classified into one type according to a plane floor plan mode.
2. The house type graph automatic processing method according to claim 1, characterized in that, the middle line refers to a longer middle line inside an elongated rectangle when the wall body is used as the elongated rectangle; further comprising: deleting the area with the area smaller than a certain threshold value from the floor plan; the certain threshold value is 0.25 square meter in area.
3. The method of claim 1, wherein the step 103 of searching for the central lines mutually closed as a closed area comprises: and judging whether the end points of the two middle lines are smaller than a threshold value or not, and connecting the head and tail end points of the middle lines in one closed area to be smaller than the threshold value in sequence.
4. The method for automatically processing a house type graph according to claim 1, wherein the step S204 further comprises: determining the floor where the room in each type of set is located; in step S202, the method of determining adjacent rooms includes:
selecting all contour lines of a room;
making a vertical extension line segment with a certain distance from the midpoint on the contour line to the outside of the room;
two rooms are considered to be adjacent if the end points of the vertical extension line fall inside the other room.
5. The method for automatically processing house type graph according to claim 1, further comprising: the method for automatically positioning the maximum local rectangle in the floor plan comprises the following steps:
s301, obtaining ordered vertex coordinates of the room polygon, and calculating four vertex coordinates of the maximum local rectangle of the room polygon to serve as training sample data;
s302, training a neural network model by adopting training sample data by taking the ordered vertex coordinates of the room polygon obtained in S301 as input values of a neural network and the four vertex coordinates of the maximum local rectangle as output values;
and S303, inputting the ordered vertex coordinates of the room polygon to be processed into the trained neural network model in S302 to obtain the four vertex coordinates of the maximum local rectangle.
6. The house type graph automatic processing method according to claim 5, characterized in that, in S301, the ordered vertex coordinates of the room polygon need to be translated, so that the plane coordinate value of one vertex is the origin, and after the four vertex coordinates of the largest local rectangle are obtained in S302, the translation direction is opposite to the translation direction of the ordered vertex coordinates of the room polygon, and the distance is the same;
the neural network model comprises:
a first layer for receiving the resulting ordered vertex coordinate data of the room polygon;
the second layer is connected with each layer and used for processing the numerical value output by the first layer, and the number of unit nodes of the second layer is the same as the number of vertexes of the room polygon;
the hidden layer is connected with the second layer and used for processing the numerical value output by the second layer;
the output layer is connected with the hidden layer and used for processing the data of the hidden layer and outputting a result;
the neural network model is a BP neural network.
7. The method for automatically processing house type graph according to claim 1, further comprising: a method for automatically locating an approximate center point of a room in a floor plan, comprising the steps of:
s401, obtaining ordered vertex coordinates of a room polygon, and calculating approximate center point coordinates of the room polygon to serve as training sample data;
s402, training a neural network model by adopting training sample data by taking the ordered vertex coordinates of the room polygon obtained in S401 as input values of a neural network and the approximate central point coordinates as output values;
and S403, inputting the ordered vertex coordinates of the room polygon to be processed into the trained neural network model in S402 to obtain the approximate center point coordinates.
8. The house type graph automatic processing method according to claim 7, characterized in that in S401, the ordered vertex coordinates of the room polygon need to be translated, so that the plane coordinate value of one vertex is the origin, and after the approximate center point coordinates are obtained in S402, the translation needs to be performed in the reverse direction, the translation direction is opposite to the translation direction of the ordered vertex coordinates of the room polygon, and the distances are the same; the neural network model comprises:
a first layer for receiving the resulting ordered vertex coordinate data of the room polygon;
the second layer is connected with each layer and used for processing the numerical value output by the first layer, and the number of unit nodes of the second layer is the same as the number of vertexes of the room polygon;
the hidden layer is connected with the second layer and used for processing the numerical value output by the second layer;
and the output layer is connected with the hidden layer and used for processing the data of the hidden layer and outputting the result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911273378.8A CN111179412B (en) | 2019-12-12 | 2019-12-12 | Automatic processing method and system for house type graph |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911273378.8A CN111179412B (en) | 2019-12-12 | 2019-12-12 | Automatic processing method and system for house type graph |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111179412A CN111179412A (en) | 2020-05-19 |
| CN111179412B true CN111179412B (en) | 2022-07-08 |
Family
ID=70655447
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911273378.8A Active CN111179412B (en) | 2019-12-12 | 2019-12-12 | Automatic processing method and system for house type graph |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111179412B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111985036B (en) * | 2020-08-27 | 2021-11-09 | 贝壳找房(北京)科技有限公司 | House type frame line drawing method and device, storage medium and electronic equipment |
| CN112417558B (en) * | 2020-11-19 | 2022-02-22 | 贝壳找房(北京)科技有限公司 | Wall decoration scheme determination method and device and computer readable storage medium |
| CN113434945A (en) * | 2021-06-30 | 2021-09-24 | 杭州群核信息技术有限公司 | Method and device for generating house type scheme |
| CN113656877B (en) * | 2021-08-23 | 2024-04-16 | 深圳须弥云图空间科技有限公司 | Multi-layer house type model generation method, device, medium and electronic equipment |
| CN115048000A (en) * | 2022-07-05 | 2022-09-13 | 厦门知本家科技有限公司 | System and method for measuring and generating house type graph based on shaft network |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108763606A (en) * | 2018-03-12 | 2018-11-06 | 江苏艾佳家居用品有限公司 | A kind of floor plan element extraction method and system based on machine vision |
| CN109801361A (en) * | 2018-12-28 | 2019-05-24 | 南京维伍网络科技有限公司 | A kind of house type editor based on illusory engine 4 and three-dimensional scenic generate system |
| CN110020502A (en) * | 2019-04-18 | 2019-07-16 | 广东三维家信息科技有限公司 | The generation method and device of floor plan |
-
2019
- 2019-12-12 CN CN201911273378.8A patent/CN111179412B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108763606A (en) * | 2018-03-12 | 2018-11-06 | 江苏艾佳家居用品有限公司 | A kind of floor plan element extraction method and system based on machine vision |
| CN109801361A (en) * | 2018-12-28 | 2019-05-24 | 南京维伍网络科技有限公司 | A kind of house type editor based on illusory engine 4 and three-dimensional scenic generate system |
| CN110020502A (en) * | 2019-04-18 | 2019-07-16 | 广东三维家信息科技有限公司 | The generation method and device of floor plan |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111179412A (en) | 2020-05-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111179412B (en) | Automatic processing method and system for house type graph | |
| CN109993827B (en) | Elevation view identification method for converting building drawing into three-dimensional BIM model | |
| Tran et al. | Shape grammar approach to 3D modeling of indoor environments using point clouds | |
| Dimitrov et al. | Non‐uniform B‐spline surface fitting from unordered 3D point clouds for as‐built modeling | |
| CN109974703B (en) | Method, system and device for constructing indoor navigation network | |
| JP4516957B2 (en) | Method, system and data structure for searching for 3D objects | |
| Einhorn et al. | Finding the adequate resolution for grid mapping-cell sizes locally adapting on-the-fly | |
| CN111597170B (en) | Method for building spatial semantic database from BIM model without damage | |
| Fang et al. | Planar shape detection at structural scales | |
| WO2021179593A1 (en) | Deep learning-based three-dimensional pipeline reconstruction method, system, medium, and apparatus | |
| CN110941871A (en) | Automatic labeling method and system based on room information in Revit three-dimensional model | |
| TWI425442B (en) | Method of Reconstructing Three - Dimensional Housing Model on Aeronautical Mapping System | |
| CN110744543B (en) | Improved PRM obstacle avoidance motion planning method based on UR3 mechanical arm | |
| Alam et al. | Towards automatic validation and healing of CityGML models for geometric and semantic consistency | |
| WO2023124160A1 (en) | Method, system and apparatus for automatically generating three-dimensional house layout, and medium | |
| CN108520543A (en) | A kind of method that relative accuracy map is optimized, equipment and storage medium | |
| Ren et al. | Intuitive and efficient roof modeling for reconstruction and synthesis | |
| CN111340100A (en) | Similarity calculation method of BIM (building information modeling) model | |
| Yang | Identify building patterns | |
| Perez-Perez et al. | Convolutional neural network architecture for semantic labeling structural and mechanical elements | |
| Zeng et al. | Integrating as-built BIM model from point cloud data in construction projects | |
| Xiong | Reconstructing and correcting 3d building models using roof topology graphs | |
| CN112668590A (en) | Visual phrase construction method and device based on image feature space and airspace space | |
| Perez | Semantically-rich as-built 3D modeling of the built environment from point cloud data | |
| As et al. | Composing frankensteins: Data-driven design assemblies through graph-based deep neural networks |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information |
Address after: 211100 floor 5, block a, China Merchants high speed rail Plaza project, No. 9, Jiangnan Road, Jiangning District, Nanjing, Jiangsu (South Station area) Applicant after: JIANGSU AIJIA HOUSEHOLD PRODUCTS Co.,Ltd. Address before: 211100 No. 18 Zhilan Road, Science Park, Jiangning District, Nanjing City, Jiangsu Province Applicant before: JIANGSU AIJIA HOUSEHOLD PRODUCTS Co.,Ltd. |
|
| CB02 | Change of applicant information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |