WO2022019675A1 - 시설 평면도에 포함된 기호 분석 장치 및 방법 - Google Patents
시설 평면도에 포함된 기호 분석 장치 및 방법 Download PDFInfo
- Publication number
- WO2022019675A1 WO2022019675A1 PCT/KR2021/009480 KR2021009480W WO2022019675A1 WO 2022019675 A1 WO2022019675 A1 WO 2022019675A1 KR 2021009480 W KR2021009480 W KR 2021009480W WO 2022019675 A1 WO2022019675 A1 WO 2022019675A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- facility floor
- plan
- equation
- floor plan
- area
- Prior art date
Links
- 238000004458 analytical method Methods 0.000 title claims abstract description 32
- 238000000034 method Methods 0.000 title claims description 28
- 238000003062 neural network model Methods 0.000 claims abstract description 37
- 238000002372 labelling Methods 0.000 claims abstract description 30
- 238000003709 image segmentation Methods 0.000 claims abstract description 10
- 230000008859 change Effects 0.000 claims description 21
- 230000015654 memory Effects 0.000 claims description 13
- 238000013528 artificial neural network Methods 0.000 claims description 11
- 238000001514 detection method Methods 0.000 claims description 11
- 238000004590 computer program Methods 0.000 claims description 6
- 230000009471 action Effects 0.000 claims description 3
- 238000013527 convolutional neural network Methods 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 12
- 230000006870 function Effects 0.000 description 12
- 238000012545 processing Methods 0.000 description 8
- 238000004891 communication Methods 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 6
- 239000003086 colorant Substances 0.000 description 4
- 238000006243 chemical reaction Methods 0.000 description 3
- 230000014509 gene expression Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 238000003491 array Methods 0.000 description 1
- 239000000470 constituent Substances 0.000 description 1
- 230000009545 invasion Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000002194 synthesizing effect Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/40—Document-oriented image-based pattern recognition
- G06V30/42—Document-oriented image-based pattern recognition based on the type of document
- G06V30/422—Technical drawings; Geographical maps
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/90—Dynamic range modification of images or parts thereof
- G06T5/94—Dynamic range modification of images or parts thereof based on local image properties, e.g. for local contrast enhancement
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/18—Extraction of features or characteristics of the image
- G06V30/1801—Detecting partial patterns, e.g. edges or contours, or configurations, e.g. loops, corners, strokes or intersections
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/18—Extraction of features or characteristics of the image
- G06V30/18105—Extraction of features or characteristics of the image related to colour
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/19—Recognition using electronic means
- G06V30/191—Design or setup of recognition systems or techniques; Extraction of features in feature space; Clustering techniques; Blind source separation
- G06V30/19173—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
Definitions
- the present invention relates to a symbol analysis apparatus and method included in a facility floor plan.
- Korean Patent Publication No. 10-1638378 (3D automatic three-dimensional modeling based on 2D drawings) methods and programs) exist.
- Korean Patent Publication No. 10-1638378 can provide highly reliable 3D spatial information because 3D modeling is performed based on spatial information and numerical values presented in a 2D plan view.
- the symbols for the window and door are both in a rectangular shape.
- the bounding box had to be manually specified and then labeled, and in addition, the bounding box cannot be rotated. There was a limit to labeling only in the shape of a rectangle.
- the problem to be solved in the embodiment of the present invention is to automatically classify the door and window symbols on the floor plan, and to label the separated areas of the door and window in units of pixels, thereby improving the learning accuracy for distinguishing the window and the door in the floor plan. to provide technology.
- a preference analysis apparatus included in a facility floor plan includes at least one memory for storing instructions for performing a predetermined operation; and one or more processors configured to be operatively connected with the one or more memories and configured to execute the instructions, wherein the operations performed by the processors include: acquiring a plurality of facility floor plans; detecting a rectangle included in each of the plurality of facility floor plans and an arc connected to the rectangle; specifying a window area and a door area based on the rectangle and the arc; labeling pixels of the specified window region with a class of a window, and labeling pixels of the specified door region with a door class; and inputting the plurality of facility floor plans and labeled data in units of pixels to a neural network model designed based on a predetermined image segmentation algorithm, and the correlation between the classes of windows and doors included in the plurality of facility floor plans and the positions of the labeled pixels By learning the weights of the neural network model to derive
- the detecting operation may include converting all parts except black included in the facility floor plan to white; and detecting the rectangle and the arc based on an outline connecting the black line segment or the white area border.
- the detecting operation may include removing the characters included in the facility floor plan.
- a pixel whose RGB information is (0, 0, 0) of a pixel included in the facility floor plan maintains RGB information
- RGB information of a pixel included in the facility floor plan is (0, 0, It may include an operation of converting RGB information of pixels other than 0) into (255, 255, 255).
- the operation of specifying the window area and the door area may include: detecting a first rectangle connected to the arc from among the rectangles as the door area; and detecting a second rectangle not connected to the arc among the rectangles as a window area.
- the first rectangle is used as the door area can be detected.
- the operation of specifying the window area and the door area may include removing from the detection when an area of the rectangle is smaller than a preset value or greater than a preset value.
- the labeling operation may include labeling pixels of all regions other than the window and the door with a class of null.
- the detecting may include: generating a first floor plan in which characters are removed from the facility floor plan through an OCR detection algorithm; generating a second plan view obtained by converting pixel information of the first plan view through Equations 1 and 2 below;
- a fourth plan view in which pixels constituting the first plan view having a color element value of 0 or more and 30 or less, a saturation element value of 80 or more and 220 or less, and a brightness element value of 150 or more and 225 or less are converted to white creating an action; generating a fifth plan view in which the black area of the third plan view and the white area of the fourth plan view are applied to the first plan view; generating a sixth plan view obtained by converting pixel information of the fifth plan view through Equations 3 to 5;
- the detecting may include: generating a first floor plan in which characters are removed from the facility floor plan through an OCR detection algorithm; generating a seventh plan view obtained by converting pixel information of the second plan view through Equations 6 to 8;
- the contour corresponding to the block hull among the approximated contours by approximating the contour based on the Douglas-Peucker algorithm, and detecting the case where the area formed by the convex hull is within a predetermined range as an arc.
- the operation of generating the neural network model sets the plurality of facility floor plans to be input to the input layer of the neural network designed based on the Mask R-CNN algorithm, and the window and door classes included in the plurality of facility floor plans are input to the output layer. and learning the weight of the neural network for deriving a correlation between the classes of windows and doors included in the plurality of facility floor plans and the positions of the labeled pixels by setting the positions of the labeled pixels to be input. have.
- An embodiment of the present invention may include a device including a neural network model generated by the device.
- a method performed by an apparatus for analyzing a preference included in a facility floor plan includes acquiring a plurality of facility floor plans; detecting a rectangle included in each of the plurality of facility floor plans and an arc connected to the rectangle; specifying a window area and a door area based on the rectangle and the arc; labeling pixels of the specified window region with a class of a window, and labeling pixels of the specified door region with a door class; And by inputting the plurality of facility floor plans and labeled data in units of pixels to a neural network model designed based on a predetermined image segmentation algorithm, the class of windows and doors included in the plurality of facility floor plans and the labeled pixels By learning the weight of the neural network model for deriving the correlation of positions, generating a neural network model that determines the positions and classes of windows and doors included in the facility floor plan based on the correlation.
- 1 is an exemplary view of a facility plan view.
- FIG. 2 is a functional block diagram of a symbol analysis apparatus included in a facility plan view according to an embodiment of the present invention.
- 3 to 5 are exemplary diagrams of an operation of detecting and labeling a door and a window by a symbol analysis apparatus included in a facility floor plan according to an embodiment of the present invention through conversion to the facility floor plan.
- FIG. 6 is an exemplary diagram illustrating a result of dividing a door and a window from a facility floor plan by a neural network model generated by a symbol analysis apparatus included in a facility floor plan according to an embodiment of the present invention.
- FIG. 7 is a 3D model performed by performing 3D modeling from a 2D plan view using a neural network model generated by a symbol analysis apparatus included in a facility floor plan according to an embodiment of the present invention and the technique of Korean Patent Publication No. 10-1638378; It is an exemplary view of the drawing.
- FIG. 8 is a flowchart of a symbol analysis method included in a facility floor plan according to an embodiment of the present invention.
- a component when it is mentioned that a component is connected or connected to another component, it may be directly connected or connected to the other component, but it should be understood that another component may exist in the middle.
- 1 is an exemplary view of a facility plan view.
- both the symbols of the window and the door include symbols of a rectangular shape.
- the bounding box can be labeled only in the form of a non-rotatable rectangle.
- the door and window symbols are arranged diagonally on the floor plan or when the door and window symbols are adjacent to other symbols, it is impossible to accurately label them, so there is a problem in that it is difficult to improve the accuracy of neural network learning.
- An embodiment of the present invention proposes a technique for automatically discriminating a door and a window symbol on a floor plan and labeling the separated door and window areas in pixel units to improve the learning accuracy for distinguishing a window from a door on a floor plan.
- FIG. 2 is a functional block diagram of a symbol analysis apparatus 100 included in a facility plan view according to an embodiment of the present invention.
- a symbol analysis apparatus 100 included in a facility plan view includes a memory 110 , a processor 120 , an input interface 130 , a display unit 140 , and a communication interface 150 .
- a memory 110 includes a processor 120 , an input interface 130 , a display unit 140 , and a communication interface 150 .
- a communication interface 150 may include
- the memory 110 may include a training data DB 111 , a neural network model 113 , and a command DB 115 .
- the learning data DB 111 may include a plurality of image files for the facility floor plan.
- the facility floor plan can be obtained through an external server, an external DB, or from an image on the Internet.
- the facility floor plan may consist of a number of pixels (ex. M*N pixels in the form of M horizontal and N vertical matrix), and each pixel is R (Red), G (Green), B (Blue). ) may include pixel information consisting of RGB element values (x, y, z) indicating the intrinsic color or HSV information indicating hue, saturation, and brightness.
- the neural network model 113 may include a neural network model that determines the class and location of the door/window symbol included in the input facility floor plan.
- the neural network model may be generated by an operation of the processor 120 to be described later and stored in the memory 110 .
- the command DB 115 may store commands capable of performing an operation of the processor 120 .
- the command DB 115 may store computer code for performing operations corresponding to operations of the processor 120 to be described later.
- the processor 120 controls the overall operation of the components included in the symbol analysis device 100 included in the facility floor plan, the memory 110 , the input interface 130 , the display unit 140 , and the communication interface 150 .
- the processor 120 may include a preference determination module 121 , a labeling module 123 , a learning module 125 , and a control module 127 .
- the processor 120 may execute the instructions stored in the memory 110 to drive the preference determination module 121 , the labeling module 123 , the learning module 125 , and the control module 127 , and the preference determination module 121 ), the labeling module 123 , the learning module 125 , and the operation performed by the control module 127 may be understood as operations performed by the processor 120 .
- the preference determination module 121 may specify a window area and a door area included in the facility floor plan for each of the plurality of facility floor plans included in the learning data DB 111 .
- the symbol identification module 121 detects and removes characters included in the facility floor plan based on an algorithm for detecting characters (ex. OCR detection algorithm), and sets all parts except black among colors included in the facility floor plan to white can be converted
- the symbol identification module 121 maintains RGB information for pixels whose RGB information is (0, 0, 0) of pixels included in the facility floor plan, and the RGB information of pixels included in the facility floor plan is (0, 0,
- An operation of converting RGB information of a pixel other than 0, 0) into (255, 255, 255) may be performed.
- the symbol discrimination module 121 may perform an operation of detecting a rectangle and an arc based on an outline connecting a black line segment or a white area border.
- the symbol determination module 121 may determine that the rectangle is not a door or window but a rectangle, and remove the corresponding rectangle from the detection target of the door or window.
- the preset value may determine the range based on the width of the door or window symbol included in the facility floor plan to be used for learning, and since the area of the symbol may vary according to the image size of the facility floor plan, It can be determined based on the image size.
- the symbol discrimination module 121 is configured to have a line segment connected to the first rectangle and forming a vertical line, in which case the arc is connected to the end of the first rectangle and the end of the line segment If so, the first rectangle can be determined as the door area.
- the symbol identification module 121 may determine a second rectangle that is not connected to an arc among the rectangles as the window area.
- the operation of the above-described preference determination module 121 may be appropriate when the facility floor plan corresponds to a high-quality original.
- the operation of the preference determination module 121 to detect the door and window from the facility floor plan containing a lot of noise will be described in more detail later with FIGS. 3 to 5 .
- the labeling module 123 may label the pixels of the window area specified by the preference determination module 121 as the class of the window, and may label the pixels of the door area specified by the preference determination module 121 as the door class. . Also, the labeling module 123 may label pixels of all areas other than windows and doors with a class of null. The labeling module 123 applies RGB information for the same color to the pixel area corresponding to the same class, and applies RGB information for different colors to the pixel area for pixel areas corresponding to different classes. can be labeled in this way.
- the labeling module 123 changes the pixel information in the facility floor plan to yellow for the pixel area where the window is located, red for the pixel area where the door is located, and black for the other pixel areas, so that each class (eg windows, doors, other areas) can be labeled.
- each class eg windows, doors, other areas
- the learning module 125 inputs a plurality of facility floor plans and labeled data in units of pixels to a neural network model designed based on an image segmentation algorithm, and correlates the classes and labeled pixel positions of windows and doors included in the plurality of facility floor plans. By learning the weight of the derived neural network model, it is possible to generate a neural network model that determines the location and class of windows and doors included in the facility floor plan based on the correlation.
- the learning module 125 sets a plurality of facility floor plans to be input to the input layer of the neural network designed based on the Mask R-CNN algorithm among the image segmentation algorithms, and to the output layer of the window/door/background included in the plurality of facility floor plans.
- the operation of learning the weight of the neural network for deriving a correlation between the class of the window and the door included in the plurality of facility floor plans and the position of the labeled pixel may be performed.
- the control module 127 can specify the area of the door and window by inputting the facility floor plan to the neural network model on which the learning has been completed, and information and numerical values about the space presented in the two-dimensional floor plan through the technology of Korean Patent Publication No. 10-1638378 By performing three-dimensional modeling based on , three-dimensional spatial information about a two-dimensional facility floor plan can be provided.
- the input interface 130 may receive a facility floor plan to be used for learning or detection.
- the display unit 140 may include a hardware configuration for outputting an image including a display panel.
- the communication interface 150 communicates with an external device to transmit/receive information.
- the communication interface 150 may include a wireless communication module or a wired communication module.
- 3 to 5 are exemplary diagrams of an operation of detecting and labeling a door and a window by a symbol analysis apparatus included in a facility floor plan according to an embodiment of the present invention through conversion to the facility floor plan.
- the symbol identification module 121 may generate a first floor plan in which characters are removed from the facility floor plan through an OCR detection algorithm.
- the symbol identification module 121 generates a second plan view obtained by converting pixel information of the first plan view through Equations 1 and 2 below to convert the first plan view into black and white colors. can do.
- the black line segment is partially cut off. This is because noise is inserted into the image of the facility floor plan so that the pixel information that was originally black is uploaded/downloaded/compressed. This is because RGB information has a value other than (0, 0, 0) due to conversion. Accordingly, the preference determination module 121 may remove noise information through an operation to be described later.
- the symbol discrimination module 121 generates a third plan view in which a portion larger than a preset area or smaller than a preset area among rectangles made of line segments constituting the second plan view is displayed in black.
- the third floor plan is used to treat an area that is very large or small compared to the area of the door and window as an outlier and remove it.
- the symbol discrimination module 121 determines that, among pixels constituting the first plan view, a color element value is 0 or more and 30 or less, a saturation element value is 80 or more and 220 or less, and a brightness element value is 150 or more.
- a fourth plan view in which pixels having a value of 225 or less are converted to white may be generated.
- the preference determination module 121 may perform a corresponding operation based on the HSV information included in the first plan view, and the HSV information may be converted and derived from the RGB information. In the fourth plan view, colors are classified based on an edge area where the color change is abrupt.
- the preference discrimination module 121 may generate a fifth plan view in which the black area of the third plan view and the white area of the fourth plan view are applied to the first plan view.
- the application of the third plan view of FIG. 3(c) is to remove an area that is very large or small compared to the area of the door and window by treating it as an outlier.
- the fourth plan view of Fig. 3(d) paying attention to the fact that noise generally occurs in the pixel area where the color change is abrupt, the color of the facility floor plan is divided into black and white based on the edge of the rapid color change, and this By synthesizing in a plan view, the effect of noise can be reduced by increasing the color contrast of the original as shown in FIG. 4(a).
- the preference discrimination module 121 may generate a sixth plan view obtained by converting pixel information of the fifth plan view through Equations 3 to 5 below.
- Equation 1 polarizes the existing color toward white and black
- Equation 4 converts the polarized color to grayscale
- Equation 5 converts polarized grayscale to black and white.
- the symbol discrimination module 121 may additionally minimize noise by applying a morphology erode operation that reduces the white part in the sixth plan view.
- the symbol identification module 121 creates a contour connecting the borders of the white area of the sixth plan view, and a rectangular area that can correspond to a window or door sign in the facility plan view can be detected.
- the symbol determination module 121 may determine a rectangle corresponding to the window or door area based on the area of the rectangle or the ratio of the width/length of the rectangle.
- the white area in Fig. 4(b) is under the following conditions may be excluded, and a rectangular area corresponding to a door or window may be detected as shown in FIG. 4(c).
- the symbol discrimination module 121 may generate a seventh plan view obtained by converting pixel information of the second plan view through Equations 6 to 8 below.
- Equation 6 may polarize the color of pixel information based on the color and saturation of the facility floor plan, Equation 7 converts the polarized color into grayscale, and Equation 8 converts grayscale into black and white.
- the symbol discrimination module 121 detects a rectangle adjacent to the arc of FIG. 5 ( b ) among the rectangles detected in FIG. 4 ( c ) as a door, and detects the remaining rectangle as a window, , the labeling module 123 may label based on the detected information.
- the learning module 125 inputs a plurality of facility floor plans ( FIG. 1 ) and data labeled in units of pixels ( FIG. 5 ( d )) to a neural network model designed based on an image segmentation algorithm, A neural network that determines the location and class of windows and doors included in a facility floor plan based on the correlation by learning the weights of a neural network model that derives a correlation between the classes and labeled pixel positions included in a plurality of facility floor plans You can create a model.
- FIG. 6 is an exemplary diagram illustrating a result of separating a door and a window from a facility floor plan by a neural network model generated by the symbol analysis apparatus 100 included in the facility floor plan according to an embodiment of the present invention.
- This neural network model can be grafted with the technology of Korean Patent Publication No. 10-1638378 as shown in FIG. 7 .
- FIG. 7 is a three-dimensional modeling from a two-dimensional plan view using the neural network model generated by the symbol analysis apparatus 100 included in the facility floor plan according to an embodiment of the present invention and the technique of Korean Patent Publication No. 10-1638378. It is an exemplary diagram of a three-dimensional drawing.
- the symbol analysis apparatus 100 included in the facility floor plan automatically distinguishes the door and window symbols on the floor plan, and labels the separated door and window itself in units of pixels in the plan view.
- Korean Patent Publication No. 10-1638378 a three-dimensional automatic three-dimensional modeling method and program based on a two-dimensional drawing), a two-dimensional plan view (Fig. 7(a) )
- it is possible to perform efficient three-dimensional modeling FIG. 7(b), FIG. 7(c) by more accurately classifying the door and window.
- FIG. 8 is a flowchart of a symbol analysis method included in a facility floor plan according to an embodiment of the present invention. Each step of the preference analysis method included in the facility floor plan according to FIG. 8 may be performed by the preference analysis apparatus 100 included in the facility floor plan described with reference to FIG. 1 , and each step is described as follows.
- the input interface 130 acquires a plurality of facility floor plans (S810).
- the symbol identification module 121 detects a rectangle included in each of a plurality of facility floor plans and an arc connected to the rectangle ( S820 ).
- the symbol identification module 121 specifies a window area and a door area based on a rectangle and an arc (S830).
- the labeling module 123 labels the pixels of the specified window area as the class of the window, and labels the pixels of the specified door area with the class of the door ( S840 ).
- the learning module 125 inputs a plurality of facility floor plans and labeled data in units of pixels to a neural network model designed based on an image segmentation algorithm, and the correlation between the classes of windows and doors included in the plurality of facility floor plans and the positions of the labeled pixels By learning the weight of the neural network model to derive , a neural network model for determining the location and class of windows and doors included in the facility floor plan is generated based on the correlation (S850).
- embodiments of the present invention may be implemented through various means.
- embodiments of the present invention may be implemented by hardware, firmware, software, or a combination thereof.
- the method according to embodiments of the present invention may include one or more Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), Digital Signal Processing Devices (DSPDs), and Programmable Logic Devices (PLDs). , FPGAs (Field Programmable Gate Arrays), processors, controllers, microcontrollers, microprocessors, and the like.
- ASICs Application Specific Integrated Circuits
- DSPs Digital Signal Processors
- DSPDs Digital Signal Processing Devices
- PLDs Programmable Logic Devices
- FPGAs Field Programmable Gate Arrays
- processors controllers
- microcontrollers microcontrollers
- microprocessors and the like.
- the method according to the embodiments of the present invention may be implemented in the form of a module, procedure, or function that performs the functions or operations described above.
- a computer program in which a software code or the like is recorded may be stored in a computer-readable recording medium or a memory unit and driven by a processor.
- the memory unit may be located inside or outside the processor, and may transmit and receive data to and from the processor by various known means.
- combinations of each block in the block diagram attached to the present invention and each step in the flowchart may be performed by computer program instructions.
- These computer program instructions may be embodied in the encoding processor of a general purpose computer, special purpose computer, or other programmable data processing equipment, such that the instructions executed by the encoding processor of the computer or other programmable data processing equipment may correspond to each block of the block diagram or
- Each step of the flowchart creates a means for performing the functions described.
- These computer program instructions may also be stored in a computer-usable or computer-readable memory that may direct a computer or other programmable data processing equipment to implement a function in a particular way, and thus the computer-usable or computer-readable memory.
- each block or each step may represent a module, segment, or part of code including one or more executable instructions for executing a specified logical function. It should also be noted that in some alternative embodiments it is also possible for the functions recited in blocks or steps to occur out of order. For example, it is possible that two blocks or steps shown one after another may in fact be performed substantially simultaneously, or that the blocks or steps may sometimes be performed in the reverse order according to the corresponding function.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Multimedia (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Image Analysis (AREA)
Abstract
Description
Claims (14)
- 시설 평면도에 포함된 기호 분석 장치에 있어서,소정의 동작을 수행하도록 하는 명령어들을 저장하는 하나 이상의 메모리; 및 상기 하나 이상의 메모리와 동작할 수 있도록 연결되어 상기 명령어들을 실행하도록 설정된 하나 이상의 프로세서를 포함하고,상기 프로세서가 수행하는 동작은,복수의 시설 평면도를 획득하는 동작;상기 복수의 시설 평면도에 각각에 포함된 직사각형 및 상기 직사각형에 연결된 호를 검출하는 동작;상기 직사각형 및 상기 호를 기초로 창문 영역과 문 영역을 특정하는 동작;상기 특정된 창문 영역의 픽셀을 창문의 클래스로 레이블링하고, 상기 특정된 문 영역의 픽셀을 문의 클래스로 레이블링하는 동작; 및소정의 이미지 분할알고리즘 기반으로 설계된 신경망 모델에 상기 복수의 시설 평면도와 픽셀 단위로 레이블링된 데이터를 입력하여, 상기 복수의 시설 평면도에 포함된 창문과 문의 클래스 및 상기 레이블링된 픽셀의 위치의 상관관계를 도출하는 상기 신경망 모델의 가중치를 학습시킴으로써, 상기 상관관계를 기초로 시설 평면도에 포함된 창문과 문의 위치 및 클래스를 판별하는 신경망 모델을 생성하는 동작을 포함하는,시설 평면도에 포함된 기호 분석 장치.
- 제1항에 있어서,상기 검출하는 동작은,시설 평면도에 포함된 검은색을 제외한 모든 부분을 흰색으로 변환시키는 동작; 및상기 검은색으로 구성된 선분 또는 흰색 영역 테두리를 연결한 윤곽을 기초로 상기 직사각형 및 상기 호를 검출하는 동작을 포함하는,시설 평면도에 포함된 기호 분석 장치.
- 제2항에 있어서,상기 검출하는 동작은,상기 시설 평면도에 포함된 문자를 제거하는 동작을 더 포함하는,시설 평면도에 포함된 기호 분석 장치.
- 제2항에 있어서,상기 흰색으로 변환시키는 동작은,시설 평면도에 포함된 픽셀의 RGB 정보가 (0, 0, 0)인 픽셀은 RGB 정보를 유지하고, 시설 평면도에 포함된 픽셀의 RGB 정보가 (0, 0, 0)이 아닌 픽셀의 RGB 정보를 (255, 255, 255)로 변환시키는 동작을 포함하는,시설 평면도에 포함된 기호 분석 장치.
- 제1항에 있어서,상기 창문 영역과 문 영역을 특정하는 동작은,상기 직사각형 중 상기 호와 연결된 제1 직사각형을 문 영역으로 검출하는 동작; 및상기 직사각형 중 상기 호와 연결되어 있지 않은 제2 직사각형을 창문 영역으로 검출하는 동작을 포함하는,시설 평면도에 포함된 기호 분석 장치.
- 제5항에 있어서,상기 문 영역으로 검출하는 동작은상기 제1 직사각형과 연결되어 수직을 이루는 선분이 존재하고, 상기 호가 상기 제1 직사각형의 말단 및 상기 선분의 말단과 연결된 경우, 상기 제1 직사각형을 문 영역으로 검출하는,시설 평면도에 포함된 기호 분석 장치.
- 제5항에 있어서,상기 창문 영역과 문 영역을 특정하는 동작은,상기 직사각형 중 넓이가 기 설정된 값보다 작거나, 기 설정된 값보다 큰 경우 상기 검출에서 제거하는 동작을 더 포함하는,시설 평면도에 포함된 기호 분석 장치.
- 제1항에 있어서,상기 레이블링하는 동작은,상기 창문 및 상기 문 이외의 모든 영역의 픽셀을 null의 클래스로 레이블링하는 동작을 포함하는,시설 평면도에 포함된 기호 분석 장치.
- 제1항에 있어서,상기 검출하는 동작은,상기 시설 평면도에서 OCR 검출 알고리즘을 통해 문자를 제거한 제1 평면도를 생성하는 동작;하기 수학식 1 및 수학식 2를 통해 상기 제1 평면도의 픽셀 정보를 변환시킨 제2 평면도를 생성하는 동작;[수학식 1](src(I): 픽셀 정보의 변경 전 원소값 (x, y, z), α: 10, β: -350, dst(I): 픽셀 정보의 변경 후 원소값 (x', y', z'))[수학식 2](R: 수학식 1에서 구해진 dst(I)의 (x', y', z') 중 x', G: 수학식 1에서 구해진 dst(I)의 (x', y', z') 중 y', B: 수학식 1에서 구해진 dst(I)의 (x', y', z') 중 z', Y: 1차원 원소값)상기 제2 평면도를 구성하는 선분으로 이루어진 직사각형 중 기 설정된 넓이보다 크거나 기 설정된 넓이보다 작은 부분만 검은색으로 나타낸 제3 평면도를 생성하는 동작;상기 제1 평면도를 구성하는 픽셀 중 색상 원소값이 0 이상 30 이하이고, 채도 원소값이 80 이상 220 이하이고, 명도 원소값이 150 이상 225 이하의 값을 갖는 픽셀을 흰색으로 변환한 제4 평면도를 생성하는 동작;상기 제1 평면도에 상기 제3 평면도의 검은색 영역과 상기 제4 평면도의 흰색 영역을 적용한 제5 평면도를 생성하는 동작;하기 수학식 3 내지 수학식 5를 통해 상기 제5 평면도의 픽셀 정보를 변환시킨 제6 평면도를 생성하는 동작;[수학식 3](src(I): 픽셀 정보의 변경 전 원소값 (x, y, z), α: 3, β: -350, dst(I): 픽셀 정보의 변경 후 원소값 (x', y', z'))[수학식 4](R: 수학식 3에서 구해진 dst(I)의 (x', y', z') 중 x', G: 수학식 3에서 구해진 dst(I)의 (x', y', z') 중 y', B: 수학식 3에서 구해진 dst(I)의 (x', y', z') 중 z', Y: 1차원 원소값)[수학식 5](Y: 수학식 4에서 구해진 1차원 원소값)상기 제6 평면도의 흰색 영역 테두리를 연결한 윤곽을 생성하여 직사각형을 검출하는 동작을 포함하는,시설 평면도에 포함된 기호 분석 장치.
- 제1항에 있어서,상기 검출하는 동작은,상기 시설 평면도에서 OCR 검출 알고리즘을 통해 문자를 제거한 제1 평면도를 생성하는 동작;하기 수학식 6 내지 수학식 8을 통해 상기 제2 평면도의 픽셀 정보를 변환시킨 제7 평면도를 생성하는 동작;[수학식 6](src(I): 픽셀 정보의 변경 전 원소값 (x, y, z), α: 57, β: -12500, dst(I): 픽셀 정보의 변경 후 원소값 (x', y', z'))[수학식 7](R: 수학식 6에서 구해진 dst(I)의 (x', y', z') 중 x', G: 수학식 6에서 구해진 dst(I)의 (x', y', z') 중 y', B: 수학식 6에서 구해진 dst(I)의 (x', y', z') 중 z', Y: 1차원 원소값)[수학식 8](Y: 수학식 7에서 구해진 1차원 원소값)상기 제7 평면도의 흰색 영역 테두리를 연결한 윤곽을 생성하는 동작; 및상기 윤곽을 Douglas-Peucker 알고리즘을 기초로 근사화하여 상기 근사화된 윤곽 중 블록 껍질(Convex hull)에 해당하는 윤곽을 검출하고, 상기 블록 껍질(Convex hull)이 이루는 넓이가 소정 범위 이내인 경우를 호로 검출하는 동작을 포함하는,시설 평면도에 포함된 기호 분석 장치.
- 제1항에 있어서,상기 신경망 모델을 생성하는 동작은,Mask R-CNN 알고리즘을 기초로 설계된 신경망의 입력 레이어에 상기 복수의 시설 평면도가 입력되도록 설정하고, 출력 레이어에는 상기 복수의 시설 평면도에 포함된 창문과 문의 클래스 및 상기 레이블링된 픽셀의 위치가 입력되도록 설정하여, 상기 복수의 시설 평면도에 포함된 창문과 문의 클래스 및 상기 레이블링된 픽셀의 위치와의 상관관계를 도출하는 신경망의 가중치를 학습시키는 동작을 포함하는,시설 평면도에 포함된 기호 분석 장치.
- 제1항 내지 제11항 중 어느 한 항의 장치가 생성한 신경망 모델을 포함하는 장치.
- 시설 평면도에 포함된 기호 분석 장치가 수행하는 방법에 있어서,복수의 시설 평면도를 획득하는 단계;상기 복수의 시설 평면도에 각각에 포함된 직사각형 및 상기 직사각형에 연결된 호를 검출하는 단계;상기 직사각형 및 상기 호를 기초로 창문 영역과 문 영역을 특정하는 단계;상기 특정된 창문 영역의 픽셀을 창문의 클래스로 레이블링하고, 상기 특정된 문 영역의 픽셀을 문의 클래스로 레이블링하는 단계; 및소정의 이미지 분할알고리즘 기반으로 설계된 신경망 모델에 상기 복수의 시설 평면도와 픽셀 단위로 레이블링된 데이터를 입력하여, 상기 복수의 시설 평면도에 포함된 창문과 문의 클래스 및 상기 레이블링된 픽셀의 위치의 상관관계를 도출하는 상기 신경망 모델의 가중치를 학습시킴으로써, 상기 상관관계를 기초로 시설 평면도에 포함된 창문과 문의 위치 및 클래스를 판별하는 신경망 모델을 생성하는 단계를 포함하는,시설 평면도에 포함된 기호 분석 방법.
- 제13항의 방법을 프로세서가 수행하도록 하는 컴퓨터 판독 가능 기록매체에 저장된 컴퓨터 프로그램.
Priority Applications (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2023504673A JP2023535084A (ja) | 2020-07-23 | 2021-07-22 | 施設平面図に含まれた記号分析装置及び方法 |
CN202180059835.2A CN116490901A (zh) | 2020-07-23 | 2021-07-22 | 设施平面图中包括的标记分析装置及方法 |
EP21846177.0A EP4184351A1 (en) | 2020-07-23 | 2021-07-22 | Symbol analysis device and method included in facility floor plan |
US18/156,853 US20230222829A1 (en) | 2020-07-23 | 2023-01-19 | Symbol analysis device and method included in facility floor plan |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
KR1020200091781A KR102208694B1 (ko) | 2020-07-23 | 2020-07-23 | 시설 평면도에 포함된 기호 분석 장치 및 방법 |
KR10-2020-0091781 | 2020-07-23 |
Related Child Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US18/156,853 Continuation US20230222829A1 (en) | 2020-07-23 | 2023-01-19 | Symbol analysis device and method included in facility floor plan |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2022019675A1 true WO2022019675A1 (ko) | 2022-01-27 |
Family
ID=74239299
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/KR2021/009480 WO2022019675A1 (ko) | 2020-07-23 | 2021-07-22 | 시설 평면도에 포함된 기호 분석 장치 및 방법 |
Country Status (6)
Country | Link |
---|---|
US (1) | US20230222829A1 (ko) |
EP (1) | EP4184351A1 (ko) |
JP (1) | JP2023535084A (ko) |
KR (3) | KR102208694B1 (ko) |
CN (1) | CN116490901A (ko) |
WO (1) | WO2022019675A1 (ko) |
Families Citing this family (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR102208694B1 (ko) * | 2020-07-23 | 2021-01-28 | 주식회사 어반베이스 | 시설 평면도에 포함된 기호 분석 장치 및 방법 |
KR20230174318A (ko) | 2022-06-17 | 2023-12-28 | 단국대학교 산학협력단 | 인공지능 모델을 이용하여 건축도면으로부터 개산견적을 산출하는 방법 |
KR20230174317A (ko) | 2022-06-17 | 2023-12-28 | 단국대학교 산학협력단 | 건축도면 구조부재 인식을 위한 인공지능 모델 생성 방법 |
KR20240037716A (ko) * | 2022-09-15 | 2024-03-22 | 한양대학교 산학협력단 | 인공신경망을 이용한 창세트 정보 자동 산출 방법 및 장치 |
US11928395B1 (en) * | 2023-04-14 | 2024-03-12 | Hubstar International Limited | Floorplan drawing conversion and analysis for space management |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20160064784A (ko) * | 2014-11-28 | 2016-06-08 | 주식회사 어반베이스 | 2차원 도면에 기반한 3차원 자동 입체모델링 방법 및 프로그램 |
KR20170109927A (ko) * | 2016-03-22 | 2017-10-10 | (주) 아키드로우 | 기계 학습 알고리즘을 이용한 도어 이미지 검출 방법 및 장치 |
KR20170115738A (ko) * | 2016-04-08 | 2017-10-18 | 이주성 | 인테리어 정보 처리 장치 및 방법 |
KR20190080904A (ko) * | 2016-11-01 | 2019-07-08 | 스냅 인코포레이티드 | 이미지 내의 객체 검출을 위한 신경망 |
KR102208694B1 (ko) * | 2020-07-23 | 2021-01-28 | 주식회사 어반베이스 | 시설 평면도에 포함된 기호 분석 장치 및 방법 |
-
2020
- 2020-07-23 KR KR1020200091781A patent/KR102208694B1/ko active IP Right Grant
-
2021
- 2021-01-19 KR KR1020210007497A patent/KR20220012789A/ko active IP Right Grant
- 2021-01-19 KR KR1020210007496A patent/KR20220012788A/ko active IP Right Grant
- 2021-07-22 EP EP21846177.0A patent/EP4184351A1/en not_active Withdrawn
- 2021-07-22 WO PCT/KR2021/009480 patent/WO2022019675A1/ko active Application Filing
- 2021-07-22 CN CN202180059835.2A patent/CN116490901A/zh not_active Withdrawn
- 2021-07-22 JP JP2023504673A patent/JP2023535084A/ja not_active Withdrawn
-
2023
- 2023-01-19 US US18/156,853 patent/US20230222829A1/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20160064784A (ko) * | 2014-11-28 | 2016-06-08 | 주식회사 어반베이스 | 2차원 도면에 기반한 3차원 자동 입체모델링 방법 및 프로그램 |
KR101638378B1 (ko) | 2014-11-28 | 2016-07-11 | 주식회사 어반베이스 | 2차원 도면에 기반한 3차원 자동 입체모델링 방법 및 프로그램 |
KR20170109927A (ko) * | 2016-03-22 | 2017-10-10 | (주) 아키드로우 | 기계 학습 알고리즘을 이용한 도어 이미지 검출 방법 및 장치 |
KR20170115738A (ko) * | 2016-04-08 | 2017-10-18 | 이주성 | 인테리어 정보 처리 장치 및 방법 |
KR20190080904A (ko) * | 2016-11-01 | 2019-07-08 | 스냅 인코포레이티드 | 이미지 내의 객체 검출을 위한 신경망 |
KR102208694B1 (ko) * | 2020-07-23 | 2021-01-28 | 주식회사 어반베이스 | 시설 평면도에 포함된 기호 분석 장치 및 방법 |
Non-Patent Citations (1)
Title |
---|
OR SIU-HANG, WONG KIN-HONG, YU YING-KIN, MING MICHAEL, CHANG -YUAN: "Highly Automatic Approach to Architectural Floorplan Image Understanding & Model Generation", VMV, 1 November 2005 (2005-11-01), pages 1 - 9, XP055889622, Retrieved from the Internet <URL:https://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.138.4263&rep=rep1&type=pdf> [retrieved on 20220209] * |
Also Published As
Publication number | Publication date |
---|---|
EP4184351A1 (en) | 2023-05-24 |
US20230222829A1 (en) | 2023-07-13 |
JP2023535084A (ja) | 2023-08-15 |
KR20220012788A (ko) | 2022-02-04 |
KR20220012789A (ko) | 2022-02-04 |
KR102208694B1 (ko) | 2021-01-28 |
CN116490901A (zh) | 2023-07-25 |
KR102208694B9 (ko) | 2022-03-11 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2022019675A1 (ko) | 시설 평면도에 포함된 기호 분석 장치 및 방법 | |
US7403656B2 (en) | Method and apparatus for recognition of character string in scene image | |
WO2015174647A1 (ko) | 사용자 인증 방법, 이를 실행하는 장치 및 이를 저장한 기록 매체 | |
WO2020130309A1 (ko) | 영상 마스킹 장치 및 영상 마스킹 방법 | |
WO2020017875A1 (en) | Electronic apparatus, method for processing image and computer-readable recording medium | |
WO2019108028A1 (ko) | 포터블 피부 상태 측정 장치, 및 피부 상태 진단 및 관리 시스템 | |
WO2020027519A1 (ko) | 영상 처리 장치 및 그 동작방법 | |
WO2017099427A1 (ko) | 손가락 마디 및 지정맥 기반의 융합형 생체 인증 방법 및 그 장치 | |
WO2020032559A2 (ko) | 뉴럴 네트워크를 이용한 질병의 진단 시스템 및 방법 | |
WO2022092743A1 (ko) | 차량 번호판에서 문자를 추출하는 방법 및 상기 방법을 수행하는 번호판 문자 추출 장치 | |
WO2019132566A1 (ko) | 멀티 심도 이미지의 자동 생성 방법 | |
US20050147304A1 (en) | Head-top detecting method, head-top detecting system and a head-top detecting program for a human face | |
WO2021006482A1 (en) | Apparatus and method for generating image | |
WO2016122068A1 (ko) | 타이어 인식 방법 및 장치 | |
WO2021071118A1 (ko) | 전경 분리 장치 및 방법 | |
WO2016104990A1 (ko) | 컨텐츠 제공 장치, 디스플레이 장치 및 그 제어 방법 | |
WO2020091337A1 (ko) | 영상 분석 장치 및 방법 | |
WO2022019390A1 (ko) | 데이터 증강 기반 사물 분석 모델 학습 장치 및 방법 | |
WO2021246822A1 (ko) | 객체 이미지 개선 방법 및 장치 | |
WO2015076433A1 (ko) | 지역 미세 패턴을 이용한 얼굴 영상 분석 방법 | |
WO2024014789A1 (ko) | 차트축감지방법 및 장치 | |
WO2020032560A2 (ko) | 진단 결과 생성 시스템 및 방법 | |
WO2022019389A1 (ko) | 데이터 증강 기반 공간 분석 모델 학습 장치 및 방법 | |
WO2017034323A1 (ko) | 적응적으로 저조도를 개선하는 영상 처리 장치 및 방법, 그리고 그를 이용한 객체 검출 장치 | |
WO2021029719A1 (ko) | 영상을 변환하는 컴퓨팅 장치 및 그 동작 방법 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 21846177 Country of ref document: EP Kind code of ref document: A1 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 202180059835.2 Country of ref document: CN |
|
ENP | Entry into the national phase |
Ref document number: 2023504673 Country of ref document: JP Kind code of ref document: A |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
ENP | Entry into the national phase |
Ref document number: 2021846177 Country of ref document: EP Effective date: 20230216 |