CN115359505A - Electric power drawing detection and extraction method and system - Google Patents
Electric power drawing detection and extraction method and system Download PDFInfo
- Publication number
- CN115359505A CN115359505A CN202210672723.0A CN202210672723A CN115359505A CN 115359505 A CN115359505 A CN 115359505A CN 202210672723 A CN202210672723 A CN 202210672723A CN 115359505 A CN115359505 A CN 115359505A
- Authority
- CN
- China
- Prior art keywords
- information
- power
- target detection
- training
- connecting line
- 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.)
- Pending
Links
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
- 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
- G06V10/762—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using clustering, e.g. of similar faces in social networks
-
- 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
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
Abstract
The invention discloses a method and a system for detecting and extracting an electric power drawing, wherein the method comprises the steps of training a target detection model in advance according to a historical electric power drawing, wherein the target detection model comprises a character detection module and a graph detection model; analyzing the power drawing to be detected according to the target detection module, and extracting character information, electric element information and connecting line information contained in the power drawing; and generating the standing book data of the power drawing according to the text information, the electrical element and the connecting line information. The method and the device can realize the detection and identification of the drawing in the intelligent checking process of the terminal strip, and the terminal information and the connection information in the drawing are obtained through the detection of the target detection model, so that a data ledger of the terminal strip information in the drawing is established, and accurate data are provided for the staff to search the terminal information and the intelligent checking in the later period.
Description
Technical Field
The invention relates to the technical field of intelligent identification of electric power drawings, in particular to a method and a system for detecting and extracting an electric power drawing, and particularly relates to a method and a system for detecting and extracting a secondary wiring drawing of a transformer substation.
Background
The secondary system devices in the transformer substation are numerous, and various cables are various and comprise relay protection devices, safety automatic devices, fault recording devices, relay protection fault information system substations, network switches, intelligent terminal devices and the like. The secondary wiring in the transformer substation is very complex, and the accuracy of the secondary wiring is related to the operation safety of the power grid, so that the secondary wiring has a very important position. Such a complex wiring structure will necessarily check the loop wiring regularly. In the face of complex and huge checking, huge workload is caused to workers.
In the past, manual checking work needs two basic conditions: 1. CAD drawings (paper documents) on site; 2. and (4) actual wiring conditions on site. The staff need know the screen cabinet, terminal number, return circuit number, the cable serial number that the terminal row belongs to through the drawing to and the subtend terminal of terminal. And then comparing one by one according to the actual wiring condition. It is easy to see from the work flow that the check work of the secondary circuit is mostly based on the inherent rule to develop, have repeatability, mechanicalness, characteristic of regularity, this provides the convenient condition for the application of artificial intelligence technology.
In the above workflow, it is particularly important for drawing review, and there is a case that the manual drawing review is assumed to be subjective, and in order to avoid the case, many people are often required to review the drawings many times. In addition, the situation that cad original files cannot be found in old substations exists, the drawing is the only checking source, under the condition of the prior art, the ocr identification technology can be used for detecting the text part of the drawing, but the connection relation of terminals in the drawing cannot be judged, and the terminal serial number cannot be obtained.
An effective solution to the problems in the related art has not been proposed yet.
Disclosure of Invention
The invention provides a method and a system for detecting and extracting an electric power drawing, aiming at the problems in the related art, and the method and the system can realize the detection and identification of the drawing in the intelligent checking process of a terminal strip, and can acquire the terminal information and the connection information in the drawing through the detection of a target detection model, thereby establishing a data ledger of the terminal strip information in the drawing and providing accurate data for the staff to search the terminal information and the intelligent checking in the later period.
The technical scheme of the invention is realized as follows:
according to one aspect of the invention, a power drawing detection and extraction method is provided.
The electric power drawing detection and extraction method comprises the following steps:
pre-training a target detection model according to a historical power drawing, wherein the target detection model comprises a character detection module and a graph detection model;
analyzing the power drawing to be detected according to the target detection module, and extracting character information, electric element information and connecting line information contained in the power drawing;
and generating the standing book data of the power drawing according to the text information, the electrical element and the connecting line information.
Wherein, according to historical electric power drawing, training target detection model in advance includes: collecting a historical power drawing, analyzing the historical power drawing, and determining the drawing type and/or drawing mode corresponding to the historical power drawing; according to the drawing type and/or the drawing mode, clustering and classifying historical electric power drawings; carrying out character recognition on the classified electric power drawings through a character recognition algorithm to obtain drawing character data, and carrying out target detection on the classified electric power drawings through a target detection algorithm to obtain drawing connecting line data and drawing electric element data; and training the drawing text data, the drawing connecting line data and the drawing electrical element data based on a pre-training model to obtain text detection models and graphic detection models of different drawing types and/or drawing modes.
Optionally, when the historical electric power drawings are clustered and classified, the historical electric power drawings are clustered and classified in a k-means clustering mode, the character recognition algorithm is based on an OCR recognition algorithm, the target detection algorithm is based on a YOLO target detection algorithm, and the pre-training model is a flying oar training model.
In addition, according to the historical power drawing, the pre-training target detection model further comprises: and performing redundant cutting on the power drawing before training in the corresponding historical power drawing according to the detection model weight and a pre-training model configured in advance, and cutting the power drawing into pictures with preset sizes.
In addition, according to the target detection module, analyzing the power drawing to be detected, and extracting the text information, the electric element information and the connecting line information included in the power drawing comprises: analyzing the power drawing to be detected, determining the drawing type and/or the drawing mode corresponding to the power drawing, and selecting a corresponding target detection model according to the determined drawing type and/or the drawing mode; performing redundant cutting on the power drawing to be detected, cutting the power drawing into graphs with a preset size, extracting graph and character information in each graph by using a character detection model in a selected target detection model, and extracting graph electrical element information and graph connecting line information in each graph by using a graph detection model in the selected target detection model; and restoring the whole connecting line information of the power drawing according to the extracted graphic electrical element information and the graphic connecting line information, and combining the connecting line information with the graphic electrical element information and the graphic character information to obtain the connecting line information, the electrical element information and the character information in the power drawing to be detected.
According to another aspect of the invention, a power drawing detection and extraction system is provided.
This electric power drawing detects extraction system includes:
the model training module is used for pre-training a target detection model according to a historical power drawing, and the target detection model comprises a character detection module and a graph detection model;
the analysis and extraction module is used for analyzing the power drawing to be detected according to the target detection module and extracting character information, electric element information and connecting line information contained in the power drawing;
and the standing book generation module is used for generating electric power drawing standing book data according to the text information, the electric elements and the connecting line information.
Additionally, the model training module includes: the device comprises a historical data analysis submodule, a cluster classification submodule, a data recognition submodule and a model training submodule, wherein the historical data analysis submodule is used for collecting a historical electric power drawing, analyzing the historical electric power drawing and determining the drawing type and/or drawing mode corresponding to the historical electric power drawing; the cluster classification submodule is used for carrying out cluster classification on the historical electric power drawing according to the drawing type and/or the drawing mode; the data identification submodule is used for carrying out character identification on the classified electric power drawings through a character identification algorithm to obtain drawing character data, and carrying out target detection on the classified electric power drawings through a target detection algorithm to obtain drawing connecting line data and drawing electric element data; and the model training submodule is used for training the drawing text data, the drawing connecting line data and the drawing electrical element data based on a pre-training model to obtain text detection models and graphic detection models of different drawing types and/or drawing modes.
Optionally, when the historical electric power drawings are clustered and classified, the historical electric power drawings are clustered and classified in a k-means clustering mode, the character recognition algorithm is based on an OCR recognition algorithm, the target detection algorithm is based on a YOLO target detection algorithm, and the pre-training model is a flying oar training model.
In addition, the model training module further comprises: and the graph cutting submodule is used for performing redundant cutting on the power drawing before training in the corresponding historical power drawing according to the detection model weight and a pre-training model configured in advance, and cutting the power drawing into pictures with preset sizes.
In addition, the analysis extraction module includes: the drawing analysis submodule, the cutting extraction submodule and the data integration submodule; the drawing analysis submodule is used for analyzing the electric power drawing to be detected, determining the drawing type and/or the drawing mode corresponding to the electric power drawing, and selecting the corresponding target detection model according to the determined drawing type and/or the drawing mode; the cutting extraction submodule is used for performing redundant cutting on the power drawing to be detected, cutting the power drawing into graphs with a preset size, extracting graph and character information in each graph by using a character detection model in the selected target detection model, and extracting graph electrical element information and graph connecting line information in each graph by using a graph detection model in the selected target detection model; and the data integration sub-module is used for restoring the integral connecting line information of the power drawing according to the extracted graphic electrical element information and the graphic connecting line information, and combining the connecting line information with the graphic electrical element information and the graphic character information to obtain the connecting line information, the electrical element information and the character information in the power drawing to be detected.
Has the advantages that:
according to the invention, the terminal information and the connection information in the drawing are obtained through the detection of the target detection model, and the data ledger of the terminal row information in the drawing is established, so that the automatic retrieval of the terminal row drawing by using an artificial intelligence means is realized, the time cost of drawing retrieval of workers in checking work is greatly saved, the retrieval error caused by artificial subjective factors in the drawing retrieval is avoided, and real-time and accurate first-hand data is provided for the later checking work; in addition, the establishment of the data standing book carries out permanent drawing recording on the condition that the transformer substation which is long in the past possibly has no original file, provides accurate original data for later-stage drawing restoration, and also provides reliable data support for future secondary circuit research work.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without creative efforts.
FIG. 1 is a schematic flow chart diagram of a method for detecting and extracting an electric power drawing according to an embodiment of the invention;
fig. 2 is a block diagram of a power drawing detection and extraction system according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived from the embodiments of the present invention by a person skilled in the art, are within the scope of the present invention.
According to the embodiment of the invention, a method and a system for detecting and extracting an electric power drawing are provided.
As shown in fig. 1, a method for detecting and extracting an electrical drawing according to an embodiment of the present invention includes:
step S101, pre-training a target detection model according to a historical power drawing, wherein the target detection model comprises a character detection module and a graph detection model;
step S103, analyzing the power drawing to be detected according to the target detection module, and extracting character information, electric element information and connecting line information contained in the power drawing;
and step S105, generating power drawing standing book data according to the character information, the electric element information and the connecting line information.
During specific application, because different drawing modes can be adopted during designing the institute design drawing, consequently will appear the terminal row drawing design pattern of difference with the institute of design as the unit. If the connection relation and the content of the terminal strip in the drawing need to be extracted by a machine, different types of drawings need to be processed differently. The accuracy of drawing information extraction can be guaranteed. Therefore, when the target detection model is trained in advance, the following steps can be included:
collecting a historical power drawing, analyzing the historical power drawing, and determining the drawing type and/or drawing mode corresponding to the historical power drawing; according to the drawing type and/or the drawing mode, clustering and classifying historical electric power drawings; carrying out character recognition on the classified electric power drawings through a character recognition algorithm to obtain drawing character data, and carrying out target detection on the classified electric power drawings through a target detection algorithm to obtain drawing connecting line data and drawing electric element data; and training the drawing text data, the drawing connecting line data and the drawing electrical element data based on a pre-training model to obtain text detection models and graphic detection models of different drawing types and/or drawing modes.
Specifically, statistical classification is carried out on drawings in the transformer substation according to drawing modes and affiliated design houses, and scene classification marking and training are carried out on the drawings in different drawing modes; respectively carrying out target detection learning and semantic segmentation learning aiming at the connection modes of different drawings; and performing redundant cutting on the drawing, and marking and training characters in the drawing respectively.
When the scene marking is carried out, firstly, drawings of different design houses are collected and arranged, and the relation between the universal terminal strip and different drawings is fused (which can comprise binding drawing modes and units aiming at the drawings of different units, dynamically writing the binding result into a recognition algorithm, marking an associated management link image appearing in the drawings, and training, wherein the purpose is that once the drawings are led in the using process, the drawings can be known to be the drawings of a certain template, and a drawing analysis method corresponding to the template is called in a later drawing analysis method); and then marking the unit area in the drawing. The labeling adopts a model detection labeling method, namely, positions appearing in a drawing are labeled through rectangles, for example: marking a unit information column and a connection relation of the unit information column from drawing information by setting a starting point coordinate as (x, y), a width as w and a height as h; extracting the key contents of the marking process, including (x, y, w, h), wherein (x, y) is the coordinates of the upper left corner of the marking boundary box, w, h are the width and height of the marking boundary box respectively, and the area of the marking boundary box after calculation and expansion is (x, y-
w,(1+
)w,h+
w) wherein (x, y-
w) is the coordinate of the upper left corner, (1 +
)w,h+
w is the width and height of the expanded bounding box. Up to this point, the processing of the data sets with respect to the different scenes is complete.
And (4) carrying out label arrangement on the data sets, wherein the label arrangement is respectively a unit information column, a universal drawing connection relation corner, a special drawing A connection pin and the like. And obtaining parameters suitable for operation detection on the classified data set in a k-means clustering mode, carrying out target detection based on a YOLO target detection algorithm, training to obtain an optimal graph detection model, carrying out character recognition based on an OCR recognition algorithm, and training to obtain an optimal character detection model.
In addition, in specific application, in order to better recognize characters, before training in the corresponding historical power diagram according to the detection model weight and a pre-training model configured in advance, the power drawing can be cut redundantly and cut into pictures with a preset size.
Specifically, after a drawing is converted into a picture format, the size of the picture, namely the width and the height of the picture, can be obtained from the basic information of the picture, and most of the basic information is represented by pixels. (1) The artificial intelligent character recognition limits the size of the picture, the optimal pixel for the mobile terminal (mobile phone APP) to process the picture is 640 x 640, the picture exceeding the pixel can be reduced, and the characters are reduced in an equal-time mode when the picture is reduced. (2) The number of characters on an A4 paper is preferably controlled within 1500 characters, so that a reader can read the content of the paper in an effective visual range. That is, the ratio of the characters to the pictures is 1/1500, so that the efficiency of machine recognition can be optimized. (3) The pixels of the drawings converted into the pictures are very high, and if the pixels are not cut, all training and recognition of characters in the drawings cannot be achieved. Thus, at the training site, the input training picture is cured to 640 x 640 pixels. Meanwhile, the characters in the picture are guaranteed to meet the requirement that the characters occupy the optimal occupation ratio of the picture, namely > =1/1500.
The method comprises the following specific operations: 1. firstly, dividing the whole picture by 640 pixels, and dividing one picture into a plurality of pictures. 2. In order to avoid the problem of incomplete characters after dividing pixels, redundancy is performed on the dividing process, namely, the redundant width and the redundant height of the pixels are increased, for example: the pixels of the complete picture are 6400 × 18000. Cutting the picture by 640 pixels, wherein the width and the height of the redundant pixels are 40 and 100 respectively, the initial coordinate of the picture of which the first part is cut is (0, 0), the width is 640, the height is 640, the real coordinate of the second picture is cut is (600, 540), and the width and the height are 640 respectively; finally, a calculation formula is obtained: w0, h0 is the total width of the picture, n is the number (n > = 0) of the picture to be cut currently, W1, h1 are redundant pixels, wa and Ha are the size of each picture after cutting, from which it can be obtained: the starting position x coordinate of the pixel cutting of the single picture is as follows: x = (n × Wa) - (n × W1), terminating when x > W0; the y coordinate of the starting position of pixel cutting of a single picture is as follows: y = (n × Ha) - (n × h 1), terminating when y > h0. 3. After the redundant cutting of the picture is finished, characters in the picture are marked, and the marking is carried out by adopting a universal marking method, namely the marking mode is that the characters are circled in the picture, the determined characters are correctly input into a training set after the marking is finished, and when the picture is actually used, a PPOCrLabel tool can be adopted for carrying out semi-automatic marking so as to accelerate the marking process; 4. the data information generated by marking comprises the correct content of characters in the picture and the position coordinates (x, y, w, h) of the characters in the picture, and is temporarily stored by using a json data structure.
In addition, in specific application, when the target detection module is used for analyzing the power drawing to be detected and extracting the character information, the electrical element information and the connecting line information contained in the power drawing, the method can be realized by the following steps: analyzing the power drawing to be detected, determining the drawing type and/or the drawing mode corresponding to the power drawing, and selecting a corresponding target detection model according to the determined drawing type and/or the drawing mode; performing redundant cutting on the power drawing to be detected, cutting the power drawing into graphs with a preset size, extracting graph and character information in each graph by using a character detection model in a selected target detection model, and extracting graph electrical element information and graph connecting line information in each graph by using a graph detection model in the selected target detection model; and restoring the whole connecting line information of the power drawing according to the extracted graphic electrical element information and graphic connecting line information, and combining the connecting line information with the graphic electrical element information and the graphic text information to obtain the connecting line information, the electrical element information and the text information in the power drawing to be detected.
In practical use, the steps can be as follows: by using the Android terminal device, the Android terminal device can automatically send the picture into the image analysis queue by importing the pdf file or shooting a high-definition picture of a field drawing. And an artificial intelligent image analysis algorithm is built in the Android terminal device, and the drawing outline and the boundary are subjected to range calibration and correction by using a scene classification technology. And then, carrying out redundant cutting on the picture, recording the cutting position, and identifying the characters of the cut picture by utilizing an OCR algorithm. Including identifying content and identifying coordinates. And detecting the connection nodes of the cut pictures by a target detection means. And record the node position coordinates. Detecting coordinates of nodes obtained after cutting, identifying coordinates of characters obtained, and performing restoration calculation on the coordinates points by combining with coordinates of cutting points (because the obtained actual information is the actual information of the picture after redundant cutting, wherein the actual information comprises text content and the relative coordinates of the picture after cutting, at the moment, the relative coordinates of a single picture are converted into actual coordinates, the conversion formula is as follows, assuming that the width and the height of the complete picture are W and H, the width and the height of the cut picture are (Wn and Hn), the cut picture is the horizontal nth picture and the vertical mth picture, the redundant width and the height of the picture are W0 and H0, the identified coordinates of the characters are Xn1 and Yn1, the width and the height are Wn1 and Hn1, then the coordinates (x, y) and the width and the height (W, H) of the complete picture where the characters are located are x (n) (Wn 1) - (W0 n) = (m H) = (m 1) - (H0W) = (W = Wn) } (H = Wn/H); carrying out connecting line detection on the complete picture by using a semantic segmentation algorithm, carrying out matching operation by combining the restored word coordinates and the node coordinates, and finally determining a corresponding relation; and respectively finding out the terminal serial number, the terminal number and the loop number on two sides of the terminal in the drawing, and the related cable number and the opposite loop number connected with the cable number according to the corresponding relation.
As shown in fig. 2, an electric power drawing detection and extraction system according to an embodiment of the present invention includes:
the model training module 201 is used for training a target detection model in advance according to a historical power drawing, wherein the target detection model comprises a character detection module and a graph detection model;
the analysis and extraction module 203 is used for analyzing the power drawing to be detected according to the target detection module and extracting character information, electric element information and connecting line information contained in the power drawing;
and the standing book generating module 205 is configured to generate electric power drawing standing book data according to the text information, the electrical element and the connecting line information.
Correspondingly, in a specific application, the model training module 201 includes: the device comprises a historical data analysis submodule (not shown in the figure), a cluster classification submodule (not shown in the figure), a data identification submodule (not shown in the figure), a graph cutting submodule (not shown in the figure) and a model training submodule (not shown in the figure), wherein the historical data analysis submodule is used for collecting a historical power drawing, analyzing the historical power drawing and determining the drawing type and/or the drawing mode corresponding to the historical power drawing; the cluster classification submodule is used for carrying out cluster classification on the historical electric power drawing according to the drawing type and/or the drawing mode; the data identification submodule is used for carrying out character identification on the classified electric power drawings through a character identification algorithm to obtain drawing character data, and carrying out target detection on the classified electric power drawings through a target detection algorithm to obtain drawing connecting line data and drawing electric element data; the graph cutting submodule is used for performing redundant cutting on the power drawing before training in the corresponding historical power drawing according to the detection model weight and a pre-training model configured in advance, and cutting the power drawing into pictures with preset sizes; and the model training submodule is used for training the drawing text data, the drawing connecting line data and the drawing electrical element data based on a pre-training model to obtain text detection models and graphic detection models of different drawing types and/or drawing modes.
The analysis and extraction module 203 comprises a drawing analysis submodule (not shown), a cutting and extraction submodule (not shown) and a data integration submodule (not shown); the drawing analysis submodule is used for analyzing the power drawing to be detected, determining the drawing type and/or the drawing mode corresponding to the power drawing, and selecting a corresponding target detection model according to the determined drawing type and/or the drawing mode; the cutting extraction submodule is used for performing redundant cutting on the power drawing to be detected, cutting the power drawing into graphs with preset size, extracting graph and character information in each graph by using a character detection model in the selected target detection model, and extracting graph electrical element information and graph connecting line information in each graph by using a graph detection model in the selected target detection model; and the data integration sub-module is used for restoring the integral connecting line information of the power drawing according to the extracted graphic electrical element information and the graphic connecting line information, and combining the connecting line information with the graphic electrical element information and the graphic character information to obtain the connecting line information, the electrical element information and the character information in the power drawing to be detected.
In summary, with the aid of the technical solution of the present invention, the terminal information and the connection information in the drawing are obtained through the target detection model detection, and the data ledger of the terminal strip information in the drawing is established, so as to implement automatic retrieval of the terminal strip drawing by using an artificial intelligence means, greatly save the time cost of drawing retrieval by workers in the checking work, avoid the retrieval errors caused by artificial subjective factors in the drawing retrieval, and provide a real-time and accurate hand of data for the later checking work; in addition, the establishment of the data standing book carries out permanent drawing recording on the condition that the transformer substation which is long in the past possibly has no original file, provides accurate original data for later-stage drawing restoration, and also provides reliable data support for future secondary circuit research work.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (10)
1. A method for detecting and extracting an electric power drawing is characterized by comprising the following steps:
pre-training a target detection model according to a historical power drawing, wherein the target detection model comprises a character detection module and a graph detection model;
analyzing the power drawing to be detected according to the target detection module, and extracting character information, electric element information and connecting line information contained in the power drawing;
and generating power drawing standing book data according to the text information, the electrical element and the connecting line information.
2. The method for detecting and extracting the power drawing as claimed in claim 1, wherein the pre-training of the target detection model according to the historical power drawing comprises:
collecting a historical power drawing, analyzing the historical power drawing, and determining the drawing type and/or drawing mode corresponding to the historical power drawing;
according to the drawing type and/or the drawing mode, clustering and classifying historical electric power drawings;
carrying out character recognition on the classified electric power drawings through a character recognition algorithm to obtain drawing character data, and carrying out target detection on the classified electric power drawings through a target detection algorithm to obtain drawing connecting line data and drawing electrical element data;
and training the drawing text data, the drawing connecting line data and the drawing electrical element data based on a pre-training model to obtain text detection models and graphic detection models of different drawing types and/or drawing modes.
3. The method for detecting and extracting the power drawings according to claim 2, wherein when the historical power drawings are subjected to cluster classification, the historical power drawings are subjected to cluster classification by adopting a k-means clustering mode, the character recognition algorithm is based on an OCR (optical character recognition) algorithm, the target detection algorithm is based on a YOLO (object detection algorithm), and the pre-training model is a flight training model.
4. The method of claim 2, wherein the pre-training of the target detection model according to the historical power drawing further comprises:
and according to the detection model weight and a pre-training model configured in advance, performing redundant cutting on the power drawing before training in the corresponding historical power drawing, and cutting the power drawing into pictures with preset sizes.
5. The method as claimed in claim 2, wherein the step of analyzing the power drawing to be detected according to the target detection module to extract the text information, the electrical component information and the connection line information included in the power drawing comprises:
analyzing the power drawing to be detected, determining the drawing type and/or the drawing mode corresponding to the power drawing, and selecting a corresponding target detection model according to the determined drawing type and/or the drawing mode;
performing redundant cutting on a power drawing to be detected, cutting the power drawing into graphs with a preset size, extracting graph and character information in each graph by using a character detection model in a selected target detection model, and extracting graph electrical element information and graph connecting line information in each graph by using a graph detection model in the selected target detection model;
and restoring the whole connecting line information of the power drawing according to the extracted graphic electrical element information and graphic connecting line information, and combining the connecting line information with the graphic electrical element information and the graphic text information to obtain the connecting line information, the electrical element information and the text information in the power drawing to be detected.
6. An electric power drawing detection and extraction system, comprising:
the model training module is used for pre-training a target detection model according to a historical power drawing, and the target detection model comprises a character detection module and a graph detection model;
the analysis and extraction module is used for analyzing the power drawing to be detected according to the target detection module and extracting character information, electric element information and connecting line information contained in the power drawing;
and the standing book generation module is used for generating electric power drawing standing book data according to the text information, the electric elements and the connecting line information.
7. The electrical drawing detection and extraction system of claim 6, wherein the model training module comprises: a historical data analysis sub-module, a cluster classification sub-module, a data recognition sub-module and a model training sub-module, wherein,
the historical data analysis submodule is used for collecting a historical power drawing, analyzing the historical power drawing and determining the drawing type and/or drawing mode corresponding to the historical power drawing;
the cluster classification submodule is used for carrying out cluster classification on the historical electric power drawing according to the drawing type and/or the drawing mode;
the data identification submodule is used for carrying out character identification on the classified electric power drawings through a character identification algorithm to obtain drawing character data, and carrying out target detection on the classified electric power drawings through a target detection algorithm to obtain drawing connecting line data and drawing electric element data;
and the model training submodule is used for training the drawing text data, the drawing connecting line data and the drawing electrical element data based on a pre-training model to obtain text detection models and graphic detection models of different drawing types and/or drawing modes.
8. The system for detecting and extracting power drawings as claimed in claim 7, wherein when the historical power drawings are subjected to cluster classification, the historical power drawings are subjected to cluster classification by adopting a k-means clustering mode, the character recognition algorithm is based on an OCR (optical character recognition) algorithm, the target detection algorithm is based on a YOLO (object detection algorithm), and the pre-trained model is a paddle training model.
9. The electrical drawing detection and extraction system of claim 7, wherein the model training module further comprises:
and the graph cutting submodule is used for performing redundant cutting on the power drawing before training in the corresponding historical power drawing according to the detection model weight and a pre-training model configured in advance, and cutting the power drawing into pictures with preset sizes.
10. The electrical drawing detection and extraction system of claim 7, wherein the analysis and extraction module comprises: the drawing analysis submodule, the cutting extraction submodule and the data integration submodule; wherein the content of the first and second substances,
the drawing analysis submodule is used for analyzing the power drawing to be detected, determining the drawing type and/or the drawing mode corresponding to the power drawing, and selecting a corresponding target detection model according to the determined drawing type and/or the drawing mode;
the cutting extraction submodule is used for performing redundant cutting on the power drawing to be detected, cutting the power drawing into graphs with preset size, extracting graph and character information in each graph by using a character detection model in the selected target detection model, and extracting graph electrical element information and graph connecting line information in each graph by using a graph detection model in the selected target detection model;
and the data integration sub-module is used for restoring the integral connecting line information of the power drawing according to the extracted graphic electrical element information and the graphic connecting line information, and combining the connecting line information with the graphic electrical element information and the graphic character information to obtain the connecting line information, the electrical element information and the character information in the power drawing to be detected.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210672723.0A CN115359505A (en) | 2022-06-14 | 2022-06-14 | Electric power drawing detection and extraction method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210672723.0A CN115359505A (en) | 2022-06-14 | 2022-06-14 | Electric power drawing detection and extraction method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115359505A true CN115359505A (en) | 2022-11-18 |
Family
ID=84030351
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210672723.0A Pending CN115359505A (en) | 2022-06-14 | 2022-06-14 | Electric power drawing detection and extraction method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115359505A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116229498A (en) * | 2023-04-24 | 2023-06-06 | 华联世纪工程咨询股份有限公司 | Automatic recognition method for column reinforcement information |
-
2022
- 2022-06-14 CN CN202210672723.0A patent/CN115359505A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116229498A (en) * | 2023-04-24 | 2023-06-06 | 华联世纪工程咨询股份有限公司 | Automatic recognition method for column reinforcement information |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109840519B (en) | Self-adaptive intelligent bill identification and input device and application method thereof | |
| CN111428723B (en) | Character recognition method and device, electronic equipment and storage medium | |
| CN108805076B (en) | Method and system for extracting table characters of environmental impact evaluation report | |
| CN111476210B (en) | Image-based text recognition method, system, device and storage medium | |
| CN108648169A (en) | The method and device of high voltage power transmission tower defects of insulator automatic identification | |
| CN113095434A (en) | Target detection method and device, electronic equipment and storage medium | |
| CN112766255A (en) | Optical character recognition method, device, equipment and storage medium | |
| CN114419647A (en) | Table information extraction method and system | |
| CN111401322A (en) | Station entering and exiting identification method and device, terminal and storage medium | |
| CN113033543A (en) | Curved text recognition method, device, equipment and medium | |
| CN113205047A (en) | Drug name identification method and device, computer equipment and storage medium | |
| CN115588202A (en) | Contour detection-based method and system for extracting characters in electrical design drawing | |
| CN115223166A (en) | Picture pre-labeling method, picture labeling method and device, and electronic equipment | |
| CN115359505A (en) | Electric power drawing detection and extraction method and system | |
| CN111507353A (en) | Chinese field detection method and system based on character recognition | |
| CN114694130A (en) | Method and device for detecting telegraph poles and pole numbers along railway based on deep learning | |
| CN113378764B (en) | Video face acquisition method, device, equipment and medium based on clustering algorithm | |
| CN114119528A (en) | Defect detection method and device for distribution line insulator | |
| CN113111728A (en) | Intelligent identification method and system for power production operation risk in transformer substation | |
| CN111340031A (en) | Equipment almanac target information extraction and identification system based on image identification and method thereof | |
| CN116541912A (en) | Method and device for generating wiring diagram based on CAD image recognition | |
| CN115953744A (en) | Vehicle identification tracking method based on deep learning | |
| CN110569831B (en) | Feature matching method and system for power equipment nameplate | |
| CN113486171A (en) | Image processing method and device and electronic equipment | |
| CN113591858A (en) | Text recognition method and device, electronic equipment and storage medium |
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 |