CN114863464A - Second-order identification method for PID drawing picture information - Google Patents
Second-order identification method for PID drawing picture information Download PDFInfo
- Publication number
- CN114863464A CN114863464A CN202210798533.3A CN202210798533A CN114863464A CN 114863464 A CN114863464 A CN 114863464A CN 202210798533 A CN202210798533 A CN 202210798533A CN 114863464 A CN114863464 A CN 114863464A
- Authority
- CN
- China
- Prior art keywords
- pid
- classification
- basic geometric
- template
- image
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 35
- 238000004364 calculation method Methods 0.000 claims abstract description 29
- 238000000605 extraction Methods 0.000 claims abstract description 3
- 238000013145 classification model Methods 0.000 claims description 37
- 238000012549 training Methods 0.000 claims description 13
- 239000013598 vector Substances 0.000 claims description 13
- 238000002372 labelling Methods 0.000 claims description 9
- 238000012216 screening Methods 0.000 claims description 7
- 238000013135 deep learning Methods 0.000 abstract description 6
- 230000008447 perception Effects 0.000 abstract description 2
- 238000010801 machine learning Methods 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 5
- 230000009286 beneficial effect Effects 0.000 description 4
- 238000001514 detection method Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 230000004075 alteration Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000006467 substitution reaction Methods 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
- 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
- G06V30/19007—Matching; Proximity measures
- G06V30/19013—Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching
-
- 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
- G06V30/19147—Obtaining sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06V30/24—Character recognition characterised by the processing or recognition method
- G06V30/248—Character recognition characterised by the processing or recognition method involving plural approaches, e.g. verification by template match; Resolving confusion among similar patterns, e.g. "O" versus "Q"
-
- 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
- G06V30/41—Analysis of document content
- G06V30/418—Document matching, e.g. of document images
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Artificial Intelligence (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a second-order identification method of PID drawing picture information, relates to the field of image identification image classification, and particularly relates to the identification problem of PID drawing pictures. The invention provides a method for realizing accurate identification and classification of PID drawing pictures by combining traditional machine learning and deep learning strong supervision, aiming at the problem of small intra-class difference in the PID pictures, the first stage adopts a mode of normalized correlation coefficient and HOG characteristic extraction correlation calculation to realize primary classification according to the common basic geometric figure characteristics of the pictures, the second stage adopts a deep learning strong supervision mode to add key parts to the pictures with higher similarity and re-label, and improves the perception capability of the model to the intra-class difference parts to improve the classification and identification performance of the model to the pictures, thereby improving the application efficiency of the PID drawing picture information in the actual engineering and improving the digital delivery capability of enterprises.
Description
Technical Field
The invention belongs to the technical field of computer image processing and image classification, and relates to a method for identifying drawing information by a computer, in particular to a PID drawing information identification method.
Background
Drawing information identification is one of the important problems in the field of digital delivery, and currently, manual determination of classification and position of a target drawing in a drawing is generally adopted. The drawing information identification algorithm based on computer vision comprises the following steps: the method comprises the steps of a traditional target detection algorithm and deep learning, wherein the traditional target detection algorithm manually filters features without good robustness to diversity changes; the method based on deep learning has the problem of high false detection rate for the classification of the targets with complicated classification and fine difference in the image; therefore, how to combine the advantages of the traditional algorithm and deep learning to design an efficient drawing and drawing identification method, which improves the problem of classification and identification of a great variety of drawings and drawings becomes a problem to be solved urgently.
Disclosure of Invention
The invention aims to provide a second-order identification method of PID drawing picture information, and in order to achieve the aim, the invention provides the following technical scheme: a second-order identification method for PID drawing picture information comprises the following steps:
s1, determining the common basic geometric figure type of the PID drawing;
s2, preparing a basic geometric figure template for the drawing, and calculating HOG characteristics of the template;
s3, performing normalized correlation coefficient matching calculation on the prepared templates one by one on the PID drawings, performing HOG characteristic calculation on the calculated result to obtain a matching result, grouping the drawing coordinate sets according to the common basic geometric figure types of the drawings, and finishing primary classification of the drawings at the first stage;
s4, extracting the picture images in a mode of expanding the boundary on the PID drawing by using the picture coordinate set to obtain a picture image set grouped according to the basic geometric figure categories shared by the pictures;
s5, performing data enhancement on the extracted picture image set in a non-deformation rotation and zooming mode, and performing re-labeling on the pictures with low distinguishing degree to serve as a picture classification data set;
s6, inputting the primary classification result of the first-stage graphs after data enhancement into a corresponding strong supervision classification model according to the common basic geometric figure category for training to obtain a robust classification model set and realize the fine classification of the second-stage graphs;
and S7, inputting the PID drawing into the step S3, inputting the result obtained after the step S3 is executed into the classification model trained in the step S6, and outputting the drawing information and the drawing coordinates to realize intelligent reading of the PID drawing information.
Preferably, in step S1, the method for determining the basic geometric figure category common to the PID drawing figures includes: the method comprises the steps of determining simple geometric figures shared among similar figures in all figures of the PID drawing, taking the shared simple geometric figures as basic geometric figures, and determining the types of the basic geometric figures in all the figures. The common basic geometry is a regular geometry, such as: round, square, triangular, rectangular, etc.; different pieces may have the same basic geometry and the pieces having the same basic geometry are classified into the same category as shown in fig. 2.
Preferably, in step S2, a basic geometric figure template is prepared for the drawing, and HOG feature calculation is performed on the template, including:
and S21, preparing basic geometric figure templates for the drawings, wherein the basic geometric figures are regular geometric figures, selecting the basic geometric figures required by the current drawings as the templates, and binarizing the template images. Because the times of the HOG characteristics of a single image are related to the number of channels of the image, the template image is compressed into a single-channel image after binarization, which is beneficial to reducing the complexity of HOG characteristic calculation;
s22, HOG feature calculation is carried out on the graph template, the gradient of each pixel of the image is calculated, the gradient comprises the size and the direction, the shape information of the graph is further captured, the image is divided into 8 × 8 small units, and the gradient histogram of each small unit is counted to serve as a feature descriptor of each small unit;
and S23, sub-connecting the feature descriptors of all the small units in the drawing template to obtain HOG feature descriptors of the drawing template, wherein the HOG feature descriptors are used for identifying feature vectors A of the drawing template.
Preferably, in step S3, the prepared templates are subjected to normalized correlation coefficient matching calculation on the PID drawings one by one, and the calculated result is subjected to HOG feature calculation, so as to obtain a matching result and complete the preliminary classification of the first-stage drawings according to a drawing coordinate set grouped by the types of basic geometric figures shared by the drawings, which is characterized by comprising:
s31, respectively carrying out normalized correlation coefficient matching on the PID drawing by using the drawing template prepared in the step S2, wherein the correlation coefficient alpha represents the similarity degree between the calculated drawing template and the area in the PID drawing search, the value range of the correlation coefficient alpha is [ -1,1], when the drawing template and the PID drawing search area are completely the same, the correlation coefficient is 1, and the value is-1 when the drawing template and the PID drawing search area are completely different;
s32, after the step of S31, primarily screening candidate areas with high matching degree in the PID drawing, wherein due to the fact that intersection lines of some areas of the PID drawing and the drawings are dense, a false screening result occurs, then HOG feature calculation is carried out on the drawing image set extracted from the PID drawing, the calculation method is the same as that of S22, and a feature vector B is obtained through calculation;
s33, measuring the similarity of the feature vector A and the feature vector B of the graph template by using the Dice distance, removing the graph with the normalized correlation coefficient matched with the mismatching through threshold control to achieve the denoising effect, grouping the screening results larger than the threshold according to the common basic geometric figure types of the graphs to form a graph coordinate set, and finishing the primary classification of the graphs in the first stage.
Preferably, in step S4, the method extracts the picture images by enlarging the boundary on the PID drawing using the picture coordinate set, and obtains a picture image set grouped according to the basic geometric figure category common to the pictures, wherein the method comprises: and extracting the image of the drawing according to the coordinates and the width and height values in the step S3 in a mode of expanding the boundary to protect the edge information of the drawing, classifying the image according to the common basic geometric figure types of the drawing to form a plurality of image sets, and extracting the image from the PID drawing according to the position information and storing the image one by one to form independent images.
Preferably, in step S5, data enhancement is performed on the extracted image set of the drawing in a non-deformation rotation and scaling manner, and a re-labeling is performed on the drawing with low degree of distinction, and the re-labeling is used as a drawing classification data set, which is characterized in that: aiming at the images of the geometric shapes, enriching a data set by adopting a rotating and scaling mode, wherein the rotating angle is 0 degree, 45 degrees, 90 degrees, 135 degrees and 180 degrees, random scaling is carried out in a range with the scaling ratio of [0.5,1.5] according to the position posture of a fitting graph appearing in a real PID image, the geometric shapes are ensured not to deform in a non-deforming scaling and rotating mode, and the full learning of image characteristics is promoted during the training of a classification model; the method is characterized in that the graph with low distinguishing degree is re-labeled, because two types of graphs and similarity thereof exist in a graph data set, the difference between the types only exists in the difference between a line and a point, the classification label is directly adopted to carry out model training, the extraction of distinguishing characteristics of the graphs is not facilitated, the key part is added to the graph with high similarity in a strong supervision mode, and the re-labeling is shown in figure 3, so that the method is beneficial to the learning of model parameters and the accuracy of model classification.
Preferably, in step S6, the initial classification result of the first-stage diagram after data enhancement is input into a corresponding strong supervised classification model according to the basic geometric figure category for training, so as to obtain a robust classification model set and realize the fine classification of the second-stage diagram, and the method is characterized in that:
s61, the strong supervision classification model is established according to the basic geometric figure category, namely the graph has N common basic geometric figures, N groups exist, each group corresponds to one classification model, N times of classification model training are needed, and finally N models are obtained;
s62, the strong supervision classification model is a classification model which is respectively constructed by grouping pictures according to the common basic geometric figure classes, the number of classification classes of each model in the N classification models is set according to the N basic geometric figures, and the image data set with enhanced data is input into the classification models in batches for full training to obtain the robust strong supervision classification model.
Preferably, in step S7, the PID drawing is input to execute step S3, and the result of executing step S3 is input to the classification model trained in step S6, so as to output the drawing information and the drawing coordinates, thereby implementing intelligent reading of the PID drawing information.
The invention has the beneficial effects that: extracting the drawing information of the PID drawing by a mode combining a traditional algorithm and deep learning, realizing the primary classification of the drawing in the PID drawing by the die matching of the normalized correlation coefficient and the HOG characteristic vector in the first stage, reducing the difficulty of the model for the various classifications of the target classes, and reducing the false detection of the primary classification by adopting the die characteristic matching; in the second stage, the drawings with small intra-class difference are re-labeled by adopting a strong supervision mode, the perception capability of the model to the difference parts among the classes is improved, the fine classification capability of the model to the drawings is improved, the classification and recognition effects of the PID drawings are improved, the efficiency of analyzing and understanding the drawing information by an engineer is improved, and the digital delivery capability of an enterprise is improved.
Drawings
FIG. 1 is a flow chart of a second-order identification method of image information according to the present invention.
FIG. 2 is a schematic diagram of common basic geometry in the drawing of the present invention.
FIG. 3 is a schematic diagram of re-labeling key parts in the primary classification class of the drawing.
FIG. 4 is a schematic diagram of the modeling and reasoning process 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 by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Example (b): a second-order identification method for drawing information of a PID drawing, as shown in FIG. 1, comprises the following steps:
s1, determining the common basic geometric figure type of the PID drawing; and extracting basic geometric figures shared by the drawings for primary classification according to the drawing catalog provided by the drawings.
S2, preparing a basic geometric figure template for the drawing, and carrying out HOG feature calculation on the template:
and S21, preparing basic geometric figure templates for the drawings, wherein the basic geometric figures are regular geometric figures, selecting the basic geometric figures required by the current drawings as the templates, and binarizing the template images. Because the times of the HOG characteristics of a single image are related to the number of channels of the image, the template image is compressed into a single-channel image after binarization, which is beneficial to reducing the complexity of HOG characteristic calculation;
s22, HOG feature calculation is carried out on the graph template, the gradient of each pixel of the image is calculated, the gradient comprises the size and the direction, the profile information of the graph is further captured, the image is divided into 8 × 8 small units, and the gradient histogram of each small unit is counted to be a feature descriptor of each small unit;
and S23, sub-connecting the feature descriptors of all the small units in the drawing template to obtain HOG feature descriptors of the drawing template, wherein the HOG feature descriptors are used for identifying feature vectors A of the drawing template.
S3, carrying out normalized correlation coefficient matching calculation on the prepared templates one by one on the PID drawings, carrying out HOG characteristic calculation on the calculated result, obtaining a matching result, grouping the drawing coordinate set according to the common basic geometric figure types of the drawings, and finishing the primary classification of the first-stage drawings:
s31, performing normalized correlation coefficient matching on the PID drawing respectively from left to right and from top to bottom in a sliding window mode by using the drawing template prepared in the step S2;
s32, primarily screening out candidate areas with high matching degree in the PID drawing, then carrying out HOG feature calculation on the drawing image set screened out from the PID drawing, wherein the calculation method is the same as that of S22, and obtaining a feature vector B through calculation;
and S33, measuring the similarity of the feature vector A and the feature vector B of the graph template by using the Dice distance, removing the graph with the normalized correlation coefficient matched with the mismatching through threshold control to achieve the denoising effect, and grouping the screening results larger than the threshold according to the common basic geometric figure types of the graphs to form a graph coordinate set to finish the primary classification of the graphs at the first stage.
S4, extracting the picture image by enlarging the boundary on the PID drawing by using the picture coordinate set, and obtaining a picture image set grouped according to the basic geometric figure category shared by the pictures; and extracting the image of the picture according to the coordinates and the width and height values in the step S3 by expanding the boundary to protect the edge information of the picture, classifying the image according to the basic geometric figure, and forming a plurality of image sets, wherein the boundary is expanded by respectively expanding the width and the height by 4 pixel units, and the expanded pixel size is determined according to the minimum distance between the pictures in the PID drawing.
And S5, performing data enhancement on the extracted image set of the image in a mode of invariant rotation and zooming, and performing re-labeling on the image with low distinguishing degree to serve as an image classification data set.
Aiming at the images of the geometric shapes, a rotation and scaling mode is adopted to enrich a data set, the rotation angle is 0 degrees, 45 degrees, 90 degrees, 135 degrees and 180 degrees, random scaling is carried out within the range of [0.5,1.5] according to the position posture of a fitting graph appearing in a real PID image, deformation of the geometric shapes is ensured in a non-deformation scaling and rotation mode, and full learning of image characteristics is promoted during classification model training.
S6, inputting the primary classification result of the first-stage graphs after data enhancement into a corresponding strong supervision classification model according to the common basic geometric figure category for training to obtain a robust classification model set and realize the fine classification of the second-stage graphs; and setting the number of fine classification categories of each model in the N classification models according to the N basic geometric figures, and respectively inputting the initial classification result of the first-stage graph after data enhancement into the corresponding classification models according to the categories of the basic geometric figures for full training to obtain a robust strong supervision classification model.
S7, inputting the PID drawing into the step S3, inputting the result of the step S3 into the classification model trained in the step S6, and outputting drawing information and drawing coordinates to realize intelligent reading of the PID drawing information; and after S1-S6, obtaining a robust strong supervision classification model, inputting a new PID drawing after model deployment, firstly executing S3 to obtain a result of primary classification according to basic geometric figures, further inputting the result into the strong supervision classification model to output a fine classification result of each drawing, and realizing the second-order identification method of the PID drawing information.
Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (7)
1. A second-order identification method of PID drawing picture information is applied to identification and extraction of information in a PID drawing, and is characterized by comprising the following steps:
s1, determining the common basic geometric figure type of the PID drawing;
s2, preparing a basic geometric figure template for the drawing, and carrying out HOG feature calculation on the template;
s3, performing normalized correlation coefficient matching calculation on the prepared templates one by one on the PID drawings, performing HOG characteristic calculation on the calculated result, obtaining a matching result, grouping the drawing coordinate set according to the common basic geometric figure types of the drawings, and finishing the primary classification of the first-stage drawings;
s4, extracting the picture image by enlarging the boundary on the PID drawing by using the picture coordinate set, and obtaining a picture image set grouped according to the basic geometric figure category shared by the pictures;
s5, performing data enhancement on the extracted image set of the drawing in a non-deformation rotation and zooming mode, and performing re-labeling on the drawing with low distinguishing degree to serve as a drawing classification data set;
s6, inputting the primary classification result of the first-stage graphs after data enhancement into a corresponding strong supervision classification model according to the common basic geometric figure category for training to obtain a robust classification model set and realize the fine classification of the second-stage graphs;
and S7, inputting the PID drawing into the step S3, inputting the result obtained after the step S3 is executed into the classification model trained in the step S6, and outputting the drawing information and the drawing coordinates to realize intelligent reading of the PID drawing information.
2. The second-order identification method for the PID drawing picture information as claimed in claim 1, wherein the step S1 determines the basic geometric figure category common to the PID drawing pictures, and the method comprises: the method comprises the steps of determining simple geometric figures shared among similar figures in all figures of the PID drawing, taking the shared simple geometric figures as basic geometric figures, and determining the types of the basic geometric figures in all the figures.
3. The second-order identification method for drawing information of PID drawing according to claim 1, wherein in step S2, a basic geometric figure template is prepared for the drawing, and HOG feature calculation is performed on the template, comprising:
s21, preparing basic geometric figure templates for the drawings, wherein the basic geometric figures are regular geometric figures, selecting the basic geometric figures required by the current drawings as the templates, and binarizing template images;
s22, HOG feature calculation is carried out on the graph template, the gradient of each pixel of the image is calculated, the gradient comprises the size and the direction, the profile information of the graph is further captured, the image is divided into 8 × 8 small units, and the gradient histogram of each small unit is counted to serve as a feature descriptor of each small unit;
and S23, sub-connecting the feature descriptors of all the small units in the drawing template to obtain HOG feature descriptors of the drawing template, wherein the HOG feature descriptors are used for identifying feature vectors A of the drawing template.
4. The second-order identification method for the drawing information of the PID drawing according to claim 1, wherein in step S3, the prepared templates are used for carrying out normalized correlation coefficient matching calculation on the PID drawing one by one, and executing HOG feature calculation on the calculated result to obtain the matching result and the drawing coordinate set grouped according to the common basic geometric figure types of the drawings to finish the primary classification of the drawings in the first stage, and the method is characterized by comprising the following steps:
s31, respectively carrying out normalized correlation coefficient matching on the PID drawing by using the drawing template prepared in the step S2, wherein the correlation coefficient alpha represents the similarity degree between the calculated drawing template and the area in the PID drawing search, the value range of the correlation coefficient alpha is [ -1,1], when the drawing template and the PID drawing search area are completely the same, the correlation coefficient is 1, and the value is-1 when the drawing template and the PID drawing search area are completely different;
s32, after the step of S31, preliminarily screening out candidate areas with high matching degree in the PID drawing, carrying out HOG feature calculation on the extracted drawing image set, wherein the calculation method is the same as that of S22, and obtaining a feature vector B through calculation;
s33, measuring the similarity of the feature vector A and the feature vector B of the graph template by using the Dice distance, removing the graphs which are mismatched by the normalized correlation coefficient matching through threshold control, and grouping the screening results which are larger than the threshold according to the common basic geometric figure types of the graphs to form a graph coordinate set to finish the primary classification of the graphs at the first stage.
5. The second-order identification method for the drawing information of the PID drawing according to claim 1, wherein in step S4, drawing images of the drawing are extracted in a manner of enlarging the boundary on the PID drawing by using the drawing coordinate set, so as to obtain a drawing image set grouped according to the basic geometric figure category shared by the drawings, and the method comprises the following steps: and extracting the image of the drawing by expanding the boundary to protect the edge information of the drawing, classifying according to the basic geometric figure shared by the drawing to form a plurality of image sets, and extracting from the PID drawing according to the position information and storing one by one into independent images.
6. The second-order identification method for the drawing information of the PID drawing according to claim 1, wherein in the step S5, the data enhancement is performed on the extracted drawing image set in a non-deformation rotation and scaling mode, and the drawing with low discrimination is re-labeled and is used as a drawing classification data set, and the method is characterized by comprising the following steps: and enriching a data set by adopting a rotating and zooming mode aiming at the images with the geometric shapes, performing additional key part re-labeling on the images of the images except for the category labels set on the images, labeling classification information for each image, and distinguishing different images.
7. The second-order identification method for the drawing information of the PID drawing, according to the claim 1, in the step S6, the primary classification result of the drawing of the first stage after data enhancement is input into the corresponding strong supervised classification model for training according to the common basic geometric figure category, so as to obtain the robust classification model set, and realize the fine classification of the drawing of the second stage, and the method is characterized by comprising the following steps:
s61, establishing a strong supervision classification model according to the common basic geometric figure categories means that the graph has N common basic geometric figures, N groups exist, each group corresponds to one classification model, N times of classification model training are needed, and finally N models are obtained;
s62, the strong supervision classification model is a classification model which is respectively constructed by grouping pictures according to the common basic geometric figure classes, the number of classification classes of each model in the N classification models is set according to the N basic geometric figures, and an image data set with enhanced data is input into the classification models in batches for full training to obtain a robust strong supervision classification model set.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210798533.3A CN114863464B (en) | 2022-07-08 | 2022-07-08 | Second-order identification method for PID drawing picture information |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210798533.3A CN114863464B (en) | 2022-07-08 | 2022-07-08 | Second-order identification method for PID drawing picture information |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114863464A true CN114863464A (en) | 2022-08-05 |
| CN114863464B CN114863464B (en) | 2023-04-14 |
Family
ID=82625638
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210798533.3A Active CN114863464B (en) | 2022-07-08 | 2022-07-08 | Second-order identification method for PID drawing picture information |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114863464B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115620014A (en) * | 2022-12-06 | 2023-01-17 | 中国石油大学(华东) | Pipeline instrument flow chart information extraction method and equipment based on deep learning |
| CN116259067A (en) * | 2023-05-15 | 2023-06-13 | 济南大学 | Method for high-precision identification of PID drawing symbols |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3029535A2 (en) * | 2014-12-03 | 2016-06-08 | Rockwell Automation Technologies, Inc. | P&ID and control system synchronization |
| CN106650820A (en) * | 2016-12-30 | 2017-05-10 | 山东大学 | Matching recognition method of handwritten electrical component symbols and standard electrical component symbols |
| CN112001047A (en) * | 2020-08-19 | 2020-11-27 | 中国电子科技集团公司第十四研究所 | PMI information-based radar key part process design method |
| CN112651378A (en) * | 2021-01-08 | 2021-04-13 | 刘连英 | Method, device and medium for identifying marking information of fastener two-dimensional drawing |
| US20220171891A1 (en) * | 2019-03-25 | 2022-06-02 | Schneider Electric Systems Usa, Inc. | Automatic extraction of assets data from engineering data sources |
| CN114694165A (en) * | 2022-06-01 | 2022-07-01 | 济南大学 | Intelligent PID drawing identification and redrawing method |
-
2022
- 2022-07-08 CN CN202210798533.3A patent/CN114863464B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3029535A2 (en) * | 2014-12-03 | 2016-06-08 | Rockwell Automation Technologies, Inc. | P&ID and control system synchronization |
| CN106650820A (en) * | 2016-12-30 | 2017-05-10 | 山东大学 | Matching recognition method of handwritten electrical component symbols and standard electrical component symbols |
| US20220171891A1 (en) * | 2019-03-25 | 2022-06-02 | Schneider Electric Systems Usa, Inc. | Automatic extraction of assets data from engineering data sources |
| CN112001047A (en) * | 2020-08-19 | 2020-11-27 | 中国电子科技集团公司第十四研究所 | PMI information-based radar key part process design method |
| CN112651378A (en) * | 2021-01-08 | 2021-04-13 | 刘连英 | Method, device and medium for identifying marking information of fastener two-dimensional drawing |
| CN114694165A (en) * | 2022-06-01 | 2022-07-01 | 济南大学 | Intelligent PID drawing identification and redrawing method |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115620014A (en) * | 2022-12-06 | 2023-01-17 | 中国石油大学(华东) | Pipeline instrument flow chart information extraction method and equipment based on deep learning |
| CN115620014B (en) * | 2022-12-06 | 2023-04-07 | 中国石油大学(华东) | Pipeline instrument flow chart information extraction method and equipment based on deep learning |
| CN116259067A (en) * | 2023-05-15 | 2023-06-13 | 济南大学 | Method for high-precision identification of PID drawing symbols |
| CN116259067B (en) * | 2023-05-15 | 2023-09-12 | 济南大学 | Method for high-precision identification of PID drawing symbols |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114863464B (en) | 2023-04-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107316031B (en) | Image feature extraction method for pedestrian re-identification | |
| CN114863464B (en) | Second-order identification method for PID drawing picture information | |
| CN106933816A (en) | Across camera lens object retrieval system and method based on global characteristics and local feature | |
| CN104036284A (en) | Adaboost algorithm based multi-scale pedestrian detection method | |
| Prasad et al. | An efficient classification of flower images with convolutional neural networks | |
| Xiang et al. | Moving object detection and shadow removing under changing illumination condition | |
| CN108073940B (en) | Method for detecting 3D target example object in unstructured environment | |
| CN112464850A (en) | Image processing method, image processing apparatus, computer device, and medium | |
| CN112686248A (en) | Certificate increase and decrease type detection method and device, readable storage medium and terminal | |
| Hatipoglu et al. | A gender recognition system from facial images using SURF based BoW method | |
| CN116503622A (en) | Data acquisition and reading method based on computer vision image | |
| CN107122714B (en) | Real-time pedestrian detection method based on edge constraint | |
| Farfan-Escobedo et al. | Towards accurate building recognition using convolutional neural networks | |
| Ibrahim et al. | Facial Expression Recognition Using Aggregated Handcrafted Descriptors based Appearance Method | |
| Verleysen et al. | Recognition of sport players' numbers using fast-color segmentation | |
| Habiba et al. | hlgp: A modified local gradient pattern for image classification | |
| Shemshaki et al. | Face detection based on fuzzy granulation and skin color segmentation | |
| Loderer et al. | Optimization of LBP parameters | |
| CN111626150A (en) | Commodity identification method | |
| CN106778679B (en) | Specific crowd video identification method based on big data machine learning | |
| Al-Shemarry et al. | Identifying License Plates in Distorted Vehicle Images: Detecting Distorted Vehicle Licence Plates Using a Novel Preprocessing Methods With Hybrid Feature Descriptors | |
| Praseeda Lekshmi et al. | Analysis of facial expressions from video images using PCA | |
| Wiwatcharakoses et al. | MSER based text localization for multi-language using double-threshold scheme | |
| Carvajal-González et al. | Feature selection by relevance analysis for abandoned object classification | |
| Paul et al. | Component based face recognition using feature matching through Hu moment invariants |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20230403 Address after: Building 1409, Building 1, No. 9 Hisense Longao, Jinan Area, China (Shandong) Pilot Free Trade Zone, Jinan City, Shandong Province, 250101 Applicant after: Shandong Zhaoen Information Technology Co.,Ltd. Address before: No. 336, nanxinzhuang Road, Shizhong District, Jinan City, Shandong Province Applicant before: University of Jinan |
|
| TA01 | Transfer of patent application right |