CN111222489A - Intelligent identification method and system for optical cable cross connecting box - Google Patents
Intelligent identification method and system for optical cable cross connecting box Download PDFInfo
- Publication number
- CN111222489A CN111222489A CN202010053985.XA CN202010053985A CN111222489A CN 111222489 A CN111222489 A CN 111222489A CN 202010053985 A CN202010053985 A CN 202010053985A CN 111222489 A CN111222489 A CN 111222489A
- Authority
- CN
- China
- Prior art keywords
- image
- identification
- port
- result
- module
- 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.)
- Withdrawn
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/20—Scenes; Scene-specific elements in augmented reality scenes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/14—Image acquisition
- G06V30/148—Segmentation of character regions
- G06V30/153—Segmentation of character regions using recognition of characters or words
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Multimedia (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses an intelligent identification method of an optical cross connecting cabinet and a corresponding system thereof, wherein the method comprises the following steps: processing an image to be processed including an optical cross-connecting box port to obtain a first image; respectively segmenting a port part and a character part in the first image to obtain a plurality of port images and corresponding first position relations thereof and a plurality of character images and corresponding second position relations thereof; identifying each port image and each character image to obtain a corresponding first identification result and a corresponding second identification result; and displaying the first recognition result and the second recognition result in a matrix according to the first position relation and the second position relation. The corresponding system comprises: the device comprises an image processing module, a first segmentation module, a first identification module, a second segmentation module, a second identification module, an identification integration module and a remote transceiving module. The port state and the character state of the optical cable cross-connecting box are located and identified, so that the state of the port and the character of the optical cable cross-connecting box is quickly obtained, the inspection time of inspection personnel is shortened, and the working efficiency is improved.
Description
Technical Field
The invention relates to the field of intelligent identification, in particular to an intelligent identification method and system for an optical cross connecting cabinet.
Background
With the popularization of network optical fibers, telecommunication optical cross connecting boxes are distributed throughout streets and alleys. In order to guarantee the best experience of users, communication companies can periodically draw people to carry out optical delivery inspection. The traditional inspection mode is that the port service condition is verified one by one to the manual work on the box, but because the inside wiring of light exchange case is mixed and disorderly, the port is numerous, if investigate one by one, the personnel of patrolling and examining will be because work load is too big and can't accomplish the task of patrolling and examining in the regulation time.
Disclosure of Invention
The embodiment of the invention aims to solve the defects in the prior art.
In order to achieve the above object, in one aspect, the embodiment of the present invention discloses an intelligent identification method for an optical cross-connecting cabinet, which implements intelligent identification of the optical cross-connecting cabinet through the following steps.
Preprocessing an image to be processed including an optical cross-connection box port to obtain a first image with an ideal visualization effect; the port part in the first image is divided to obtain a plurality of port images and a first position relation of each port image in the first image; dividing character parts in the first image to obtain a plurality of character images and a second position relation of each character image in the first image; identifying each port image through a port classifier to obtain a plurality of corresponding first identification results; identifying each character image through a neural network model to obtain a plurality of corresponding second identification results; and generating a third recognition result in a matrix form according to the first position relation and the second position relation and the corresponding first recognition result and the second recognition result.
In one example, according to actual conditions, the patrol personnel corrects the first recognition result and/or the second recognition result which do not accord with the preset rule in the third recognition result, generates a fourth recognition result and uploads the fourth recognition result to the cloud for storage.
In one example, the process of the port classifier identifying each port image to obtain the corresponding plurality of first identification results may be subdivided as follows: setting a port classifier comprising a plurality of models, wherein the plurality of models comprises a first model and a second model;
identifying the port image through the first model to generate a first-order identification result; if the first-order recognition result accords with a preset rule, outputting the first-order recognition result as a first recognition result; if the first-order identification result does not accord with the preset rule, re-identifying the port image through the second model to generate a second-order identification result, and outputting a result which is close to the preset rule in the first-order identification result and the second-order identification result as a first identification result; or respectively identifying the port images through the first model and the second model to obtain two identification results, forming a result set by the two identification results, and outputting the result set as a first identification result.
On the other hand, the embodiment of the invention discloses an intelligent identification system of an optical delivery box, which comprises an image processing module, a first segmentation module, a first identification module, a second segmentation module, a second identification module, an identification comprehensive module and a remote transceiving module. Wherein the content of the first and second substances,
the image processing module is used for preprocessing an image to be processed to obtain a first image with an ideal visualization effect;
the first segmentation module is used for segmenting the port part in the first image to obtain a plurality of port images and a first position relation of each port image in the first image;
the first identification module is used for identifying each port image through the port classifier to obtain a plurality of corresponding first identification results;
the second segmentation module is used for segmenting the character part in the first image to obtain a plurality of character images and a second position relation of each character image in the first image;
the second recognition module is used for recognizing each character image through the neural network model to obtain a plurality of corresponding second recognition results;
the identification integration module is used for generating a third identification result in a matrix form according to the first position relation and the second position relation and the corresponding first identification result and the second identification result;
the remote transceiving module is used for uploading the third identification result to the cloud storage.
In one example, a check revision module may also be provided; the inspection revising module revises the third identification result by inspection personnel according to the actual situation to obtain a fourth identification result; and uploading the fourth identification result to a cloud terminal for storage through a remote transceiving module.
In one example, the recognition rate is improved by providing a plurality of recognition sub-modules in the first recognition module, wherein the plurality of recognition sub-modules includes a first sub-module and a second sub-module.
The first sub-module identifies the port image to generate a first-order identification result; if the first-order recognition result accords with a preset rule, outputting the first-order recognition result as a first recognition result; if the first-order identification result does not accord with the preset rule, the second submodule identifies the port image again to generate a second-order identification result, and outputs a result which is close to the preset rule in the first-order identification result and the second-order identification result as a first identification result; or the first sub-module and the second sub-module respectively identify the port images to obtain two identification results, the two identification results form a result set, and the result set is output as a first identification result.
The embodiment of the invention has the advantages that: the port state and the character state of the optical cross-connecting box are positioned and identified through knowledge in the aspects of machine vision, deep learning, man-machine interaction and the like, and the purpose is to quickly obtain the port state and the character state of the optical cross-connecting box through an artificial intelligence method so as to shorten the inspection time of inspection personnel and improve the working efficiency.
Drawings
Fig. 1 is a flow chart of a method for recognizing only an optical cross connecting box according to an embodiment of the present invention;
FIG. 2 is a block diagram of an optical cross-connect identification only system according to an embodiment of the present invention;
FIG. 3 is a diagram of an image to be processed including an optical cross-connect port according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of a port image matrix according to an embodiment of the invention;
FIG. 5 is a schematic diagram of a character image according to an embodiment of the present invention;
fig. 6 is a diagram illustrating a third recognition result according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, 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 some, but not all, embodiments of the present invention. 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.
Fig. 1 is a flow chart of a method for identifying only optical communication boxes according to an embodiment of the present invention, as shown in fig. 1. The method comprises the following steps:
step S110 image preprocessing
And preprocessing the image to be processed including the optical cross-connecting box port to obtain a first image with ideal visualization effect.
In actual life, inspection personnel mostly take pictures by using a mobile phone or a camera to obtain images to be processed. Due to the influence of various factors such as a shooting angle, shooting light, shooting equipment and the like, a series of interference components which are not beneficial to image identification, such as overexposure, underexposure and the like, may occur to the image to be processed. By processing the image to be processed by using a digital image processing technique, a first image with an ideal visualization effect can be obtained.
In one example, the image to be processed is dynamically compressed by adjusting the gamma curve of the image to be processed, so that the pixels with low brightness (underexposed portions) become bright and the pixels with high brightness (overexposed portions) become dark. And then processing the problem of uneven exposure of the image by fitting the information such as the brightness, the contour, the details and the like of the image to be processed.
Step S121: segmenting port images
The port portions in the first image are segmented using a checkerboard algorithm. A plurality of port images are obtained, and a first position relation of each port image in the first image is obtained simultaneously.
According to the characteristic that the images of the optical cross-connect box are arranged in a port matrix mode, the first image can be segmented by using a chessboard algorithm, and the port image and the position relation of the port image in the first image are obtained. Considering that in a first image containing ports of an optical cross-connect box, the distribution of wiring is disordered and numerous, and meanwhile, the situation that the ports are shielded may exist, it is difficult to completely position and identify the positions of the ports to obtain port images under the condition of ensuring better universality by directly using a machine learning or deep learning method. Thus, in one embodiment, the characteristics of each port may be determined in advance, and a certain characteristic may be determined as the first characteristic. Determining the positions of ports containing the first characteristics in the first image according to the first characteristics of the ports, and segmenting the ports containing the first characteristics; and according to the determined position of the port with the first characteristic in the first image, segmenting the rest ports without the first characteristic by an adaptive chessboard algorithm.
Step S122: segmenting character images
And segmenting the character part in the first image to obtain a plurality of character images and a second position relation of each character image in the first image.
And determining the character part at the head or tail of each row in the port matrix according to the characteristic of the corresponding relation formed by the character part and the row of the port matrix in the image of the optical cross-connect box. The character image is generated by dividing the character parts into at least two layers corresponding to each other in a unit of division. And obtaining the position relation of each character image in the first image according to the corresponding relation of the character part and the row of the port matrix.
Step S131: port image recognition
And identifying each port image by using intelligent identification modes such as a trained support vector machine classifier, a convolutional neural network model and the like to obtain a first identification result corresponding to the port image.
The trained support vector machine classifier mentioned here takes a port image as a sample and takes a gradient histogram HOG of the port image as a feature, and a support vector machine SVM is used for training a feature vector of the port image. The trained convolutional neural network is obtained by training by taking the port image as a sample.
In actual use, the port image and the corresponding position relation thereof can be obtained by only using one identification mode, but considering that the problem of insufficient identification accuracy rate possibly exists when a certain identification method is singly used, more than two kinds of intelligent identification can be simultaneously carried out on the port image so as to improve the identification accuracy.
In the process of performing intelligent recognition by adopting at least more than two recognition methods, a recognition sequence can be designed. For example, firstly, an SVM-HOG port classifier is used for identifying the port image, and an identification result is directly generated for the port with accurate identification. And identifying the port with inaccurate identification based on the convolutional neural network to obtain an identification result closer to a real result, and summarizing the two results to generate a first identification result. Besides designing a specific recognition sequence and carrying out multi-order recognition, the method can also carry out recognition in different modes for the same sample for multiple times, and take multiple recognition results as a result set of a first recognition result, so that the accuracy of the multiple recognition results in the result set is confirmed manually. For example, a certain port image is used as a sample of an SVM-HOG port classifier and a CNN recognition model, and two corresponding recognition results are obtained. Both of these recognition results are output as the first recognition result, and at step S150, the two results are discriminated manually.
Step S132: character image recognition
And recognizing each character image by using the trained YOLO network model to obtain a plurality of corresponding second recognition results.
Step S140: comprehensive recognition result
And establishing a visual matrix model according to a preset rule to display the first recognition result and the second recognition result. And generating a third recognition result in a two-dimensional matrix form according to a first recognition result obtained by recognizing the port image and the first position relation thereof, and a second recognition result obtained by recognizing the character image and the second position relation thereof. Each element in the third recognition result corresponds to one first recognition result or one second recognition result.
Step S150: human-machine interaction correction
And inquiring the third identification result by the inspection personnel. And correcting the first recognition result or the second recognition result which is poorly fitted with the label of the network model in the third recognition result according to the actual situation to generate a fourth recognition result. And uploading the obtained fourth recognition result most similar to the real situation to a cloud for storage.
By adopting the method, the positioning identification of the port and the port state of the mobile optical cable optical cross connecting box can be completed through the technologies of machine learning, deep learning, man-machine interaction and the like, the inspection man-hour of inspection personnel is shortened, and meanwhile, the identification result of each time is directly transmitted to the cloud for storage.
The establishment of the system based on the method can comprise the following steps: the system comprises an image processing module, a first segmentation module, a first identification module, a second segmentation module, a second identification module, an identification integration module, an inspection revision module and a remote transceiving module. As shown in fig. 2.
The image processing module is used for processing the acquired image to be processed, and specifically aims at the image to be processed comprising the optical cross-connecting box port. The method is mainly obtained by a method that inspection personnel take pictures by a mobile phone or a camera. Due to the influence of various factors such as a shooting angle, shooting light, shooting equipment and the like, a series of interference components which are not beneficial to image identification, such as overexposure, underexposure and the like, may occur to the image to be processed. By processing the image to be processed by using a digital image processing technique, a first image with an ideal visualization effect can be obtained.
In one example, the image to be processed is dynamically compressed by adjusting the gamma curve of the image to be processed, so that the pixels with low brightness (underexposed portions) become bright and the pixels with high brightness (overexposed portions) become dark. And then processing the problem of uneven exposure of the image by fitting the information such as the brightness, the contour, the details and the like of the image to be processed.
The first segmentation module segments the port portions in the first image using a checkerboard algorithm. A plurality of port images are obtained, and a first position relation of each port image in the first image is obtained simultaneously. According to the characteristic that the images of the optical cross-connect box are arranged in a port matrix mode, the first image can be segmented by using a chessboard algorithm, and the port image and the position relation of the port image in the first image are obtained. Considering that in a first image containing ports of an optical cross-connect box, the distribution of wiring is disordered and numerous, and meanwhile, the situation that the ports are shielded may exist, it is difficult to completely position and identify the positions of the ports to obtain port images under the condition of ensuring better universality by directly using a machine learning or deep learning method. Thus, in one embodiment, the characteristics of each port may be determined in advance, and a certain characteristic may be determined as the first characteristic. Determining the positions of ports containing the first characteristics in the first image according to the first characteristics of the ports, and segmenting the ports containing the first characteristics; and according to the determined position of the port with the first characteristic in the first image, segmenting the rest ports without the first characteristic by an adaptive chessboard algorithm.
The first recognition module recognizes each port image by using intelligent recognition modes such as a trained support vector machine classifier, a convolutional neural network model and the like to obtain a first recognition result corresponding to the port image.
The trained support vector machine classifier mentioned here takes a port image as a sample and takes a gradient histogram HOG of the port image as a feature, and a support vector machine SVM is used for training a feature vector of the port image. The trained convolutional neural network is obtained by training by taking the port image as a sample.
In actual use, the port image and the corresponding position relation thereof can be obtained by only using one identification mode, but considering that the problem of insufficient identification accuracy rate possibly exists when a certain identification method is singly used, more than two kinds of intelligent identification can be simultaneously carried out on the port image so as to improve the identification accuracy.
Therefore, the first identification module can comprise a plurality of identification sub-modules to realize intelligent identification in various ways. In the process of performing intelligent recognition by adopting at least more than two recognition methods, a recognition sequence can be designed. For example, firstly, an SVM-HOG port classifier is used for identifying the port image, and an identification result is directly generated for the port with accurate identification. And identifying the port with inaccurate identification based on the convolutional neural network to obtain an identification result closer to a real result, and summarizing the two results to generate a first identification result. Besides designing a specific recognition sequence and carrying out multi-order recognition, the method can also carry out recognition in different modes for the same sample for multiple times, and take multiple recognition results as a result set of a first recognition result, so that the accuracy of the multiple recognition results in the result set is confirmed manually. For example, a certain port image is used as a sample of an SVM-HOG port classifier and a CNN recognition model, and two corresponding recognition results are obtained. Both of these recognition results are output as the first recognition result, and at step S150, the two results are discriminated manually.
The second segmentation module segments the character part in the first image to obtain a plurality of character images and a second position relation of each character image in the first image. And determining the character part at the head or tail of each row in the port matrix according to the characteristic of the corresponding relation formed by the character part and the row of the port matrix in the image of the optical cross-connect box. The character image is generated by dividing the character parts into at least two layers corresponding to each other in a unit of division. And obtaining the position relation of each character image in the first image according to the corresponding relation of the character part and the row of the port matrix.
And the second recognition module recognizes each character image by using the trained YOLO network model to obtain a plurality of corresponding second recognition results.
And the identification comprehensive module establishes a visual matrix model for displaying the identification results generated by the first identification module and the second identification module. And generating a third recognition result in a two-dimensional matrix form according to a first recognition result obtained by recognizing the port image and the first position relation thereof, and a second recognition result obtained by recognizing the character image and the second position relation thereof. Each element in the third recognition result corresponds to one first recognition result or one second recognition result.
The inspection revision module is used for providing the polling personnel with inquiry and revision of the third identification result. And the patrol personnel corrects the first recognition result or the second recognition result which is poorly fitted with the label of the network model in the third recognition result according to the actual situation to generate a fourth recognition result.
And the remote transceiving module uploads the third identification result and the fourth identification result to a cloud storage so as to be convenient for query.
In a specific embodiment, the identification result in the system is displayed through application software on mobile equipment, so that patrol personnel can conveniently and quickly know the specific situation of the optical delivery box. The system rapidly identifies the acquired image to be processed to obtain an identification result, and the inspection personnel revises errors in the identification result according to actual conditions.
The image to be processed including the port of the optical cross-connect box collected by the inspection personnel is shown in fig. 3. The light boxes therein show a total of 26 rows of 12 ports each, and a corresponding character for each row.
And preprocessing the image to be processed to obtain a first image with an ideal visualization effect.
The first image is segmented into 312(26 rows by 12 per row) port images. The 312 port images may be grouped into a matrix as shown in fig. 4 according to the position relationship of each port image in the first image. In order to omit part of the port images for the sake of convenience and description, in the following description, 20 port images specifically displayed will be also recited with emphasis. It is to be understood that other port images not specifically shown will be processed in the same manner as the specifically shown 20 port images.
In the port image matrix shown in fig. 4, the ports are covered with red cap packing rubber sleeves in the port images with numbers 1-1, 1-2, 1-12 and 25-2; in the port images numbered 2-1, 2-12, 25-1, 25-11, the ports are not normally displayed and the ports are not outlined; in port images of numbers 3-1, 3-12, 26-1 and 26-11, the ports are circular metal caps; in port images with numbers 1-11, 2-2, 2-11 and 25-12, a port is connected with a cable; in the port images with the numbers 3-2, 3-11, 26-2 and 26-12, the port outlines are clear, but no connecting end is arranged or the connecting end is damaged and exposed.
Firstly, recognizing the ports in the graph 4 by using an SVM-HOG port classifier, and generating a recognition result of a first stage. In order to avoid the lack of accuracy of the identification result, the CNN identification model is used for carrying out second identification on the port image which is poor in the fitting condition with the label in the identification result of the first stage, and the identification result of the second stage is generated. And synthesizing the recognition result of the first stage and the recognition result of the second stage to generate a first recognition result.
Specifically, for the port images numbered 1-1, 1-2, 1-12 and 25-2 in fig. 4, the port classifier determines the port images as a "red cap" state according to the fact that red cap rubber sleeves are sleeved on the ports; the port images with numbers 2-1, 2-12, 25-1 and 25-11 are judged to be in a shielding state by the port classifier according to the condition that the port is not normally displayed and the port is unclear in outline; the port images with numbers of 3-1, 3-12, 26-1 and 26-11 are judged to be in a round cap state by a port classifier according to the fact that the ports are round metal caps; the port images with the numbers of 1-11, 2-2, 2-11 and 25-12 are connected with cables according to the ports, and the port classifier judges the port images to be in a connection state; the port images with the numbers 3-2, 3-11, 26-2 and 26-12 are judged to be in a defect state by the port classifier according to the clear port contour without connecting ends or the damaged and exposed connecting ends. Therein for port images numbered 2-1, 2-12, 25-1, 25-11. The maximum probability of the 'shielding' state is caused by the shielding of the connecting line of other ports or other shielding objects when the image is acquired. The inspection personnel is required to inspect the port corresponding to the port image in the shielding state according to the field condition, and correct the shielding state in the first identification result according to the inspection result.
The character image divides the first image with 1 unit of characters corresponding to each 3 adjacent lines. The character image obtained by the segmentation is shown in fig. 5.
And identifying the characters corresponding to each 3 adjacent lines by using a YOLO network model to obtain a second identification result corresponding to each line of characters. The second recognition result is embodied in the form of text, the content of which is used to mark or illustrate the line.
A matrix model of 26 rows of 13 elements each is built. The 312 first recognition results obtained by recognizing the port images are displayed in the 26 × 13 matrix model according to the first positional relationship in which their corresponding port images are located in the first image. Specifically, the display may be performed in 26 rows, and the first 12 elements of each row correspond to positions. A second recognition result for recognizing the character image is displayed in the 26 x 13 matrix model in accordance with the second positional relationship in which the row corresponding to each character image is located. Specifically, the display can be performed in 26 rows, and the 13 th element of each row corresponds to the position. A visualization matrix model as shown in fig. 6, i.e. a third recognition result, is obtained. The visualization matrix model comprises recognition results of all port images and character images in the first image.
And inquiring the third identification result by the inspection personnel. And correspondingly searching for a port identified as 'shielded' in the real optical cross-connect according to the position of the 'shielded' state displayed in the third identification result in the matrix model, carrying out manual identification on the port, updating the third identification result according to the result of the manual identification, and generating a fourth identification result. In the process, the third recognition result and the fourth recognition result are uploaded to the cloud server for storage, and other users can conveniently inquire the contents recognized by the machine and the contents recognized by human-computer interaction.
According to the intelligent identification method and system for the optical cross-connecting box, the port state and the character state of the optical cross-connecting box are located and identified through knowledge in the aspects of machine vision, deep learning, man-machine interaction and the like, and the purpose of quickly obtaining the states of the port and the character of the optical cross-connecting box through an artificial intelligence method is achieved, so that the inspection time of inspection personnel is shortened, and the working efficiency is improved.
The above embodiments are provided to further explain the objects, technical solutions and advantages of the present invention in detail, it should be understood that the above embodiments are merely exemplary embodiments of the present invention and are not intended to limit the scope of the present invention, and any modifications, equivalents, improvements and the like made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (6)
1. An intelligent identification method for an optical cross connecting cabinet is characterized by comprising the following steps:
preprocessing an image to be processed including an optical cross-connection box port to obtain a first image with an ideal visualization effect;
dividing the port part in the first image to obtain a plurality of port images and a first position relation of each port image in the first image; dividing character parts in the first image to obtain a plurality of character images and a second position relation of each character image in the first image;
identifying each port image through a port classifier to obtain a plurality of corresponding first identification results; identifying each character image through a neural network model to obtain a plurality of corresponding second identification results;
and generating a third recognition result in a matrix form according to the first position relation and the second position relation and the corresponding first recognition result and the second recognition result.
2. The method of claim 1, further comprising: and according to the actual situation, correcting the first recognition result and/or the second recognition result which do not accord with the preset rule in the third recognition result, generating a fourth recognition result and uploading the fourth recognition result to the cloud for storage.
3. The method according to claim 1, wherein the step of identifying each port image by the port classifier to obtain a plurality of corresponding first identification results comprises:
the port classifier comprises a plurality of models including a first model and a second model;
identifying the port image through the first model to generate a first-order identification result; if the first-order recognition result accords with a preset rule, outputting the first-order recognition result as a first recognition result; if the first-order identification result does not accord with the preset rule, re-identifying the port image through the second model to generate a second-order identification result, and outputting a result which is close to the preset rule in the first-order identification result and the second-order identification result as a first identification result;
and respectively identifying the port images through the first model and the second model to obtain two identification results, forming a result set by the two identification results, and outputting the result set as a first identification result.
4. An intelligent identification system of an optical delivery box is characterized by comprising an image processing module, a first segmentation module, a first identification module, a second segmentation module, a second identification module, an identification comprehensive module and a remote transceiving module; wherein the content of the first and second substances,
the image processing module is used for preprocessing an image to be processed to obtain a first image with an ideal visualization effect;
the first segmentation module is used for segmenting the port part in the first image to obtain a plurality of port images and a first position relation of each port image in the first image;
the first identification module is used for identifying each port image through the port classifier to obtain a plurality of corresponding first identification results;
the second segmentation module is used for segmenting the character part in the first image to obtain a plurality of character images and a second position relation of each character image in the first image;
the second recognition module is used for recognizing each character image through the neural network model to obtain a plurality of corresponding second recognition results;
the identification integration module is used for generating a third identification result in a matrix form according to the first position relation and the second position relation and the corresponding first identification result and the second identification result;
the remote transceiving module is used for uploading the third identification result to the cloud storage.
5. The system of claim 4, further comprising an inspection revision module;
the inspection revising module is used for revising the third identification result by the inspection personnel according to the actual condition to obtain a fourth identification result;
and uploading the fourth identification result to a cloud terminal for storage through a remote transceiving module.
6. The system of claim 4, wherein the first identification module comprises a plurality of identification submodules, the plurality of identification submodules comprising a first submodule and a second submodule;
the first sub-module identifies the port image to generate a first-order identification result; if the first-order recognition result accords with a preset rule, outputting the first-order recognition result as a first recognition result; if the first-order identification result does not accord with the preset rule, the second submodule identifies the port image again to generate a second-order identification result, and outputs a result which is close to the preset rule in the first-order identification result and the second-order identification result as a first identification result;
the first sub-module and the second sub-module respectively identify the port images to obtain two identification results, the two identification results form a result set, and the result set is output as a first identification result.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010053985.XA CN111222489A (en) | 2020-01-17 | 2020-01-17 | Intelligent identification method and system for optical cable cross connecting box |
| CN202010582244.0A CN111507316A (en) | 2020-01-17 | 2020-06-23 | Intelligent identification method and system for optical cable cross connecting box |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010053985.XA CN111222489A (en) | 2020-01-17 | 2020-01-17 | Intelligent identification method and system for optical cable cross connecting box |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111222489A true CN111222489A (en) | 2020-06-02 |
Family
ID=70829588
Family Applications (2)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010053985.XA Withdrawn CN111222489A (en) | 2020-01-17 | 2020-01-17 | Intelligent identification method and system for optical cable cross connecting box |
| CN202010582244.0A Pending CN111507316A (en) | 2020-01-17 | 2020-06-23 | Intelligent identification method and system for optical cable cross connecting box |
Family Applications After (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010582244.0A Pending CN111507316A (en) | 2020-01-17 | 2020-06-23 | Intelligent identification method and system for optical cable cross connecting box |
Country Status (1)
| Country | Link |
|---|---|
| CN (2) | CN111222489A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2022217989A1 (en) * | 2021-04-17 | 2022-10-20 | 华为技术有限公司 | Communication port management method and related device |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106845478B (en) * | 2016-12-30 | 2019-09-10 | 同观科技（深圳）有限公司 | A kind of secondary licence plate recognition method and device of character confidence level |
| CN107832767A (en) * | 2017-11-14 | 2018-03-23 | 深圳码隆科技有限公司 | Container number identification method, device and electronic equipment |
| CN108596166B (en) * | 2018-04-13 | 2021-10-26 | 华南师范大学 | Container number identification method based on convolutional neural network classification |
| CN109815998A (en) * | 2019-01-08 | 2019-05-28 | 科大国创软件股份有限公司 | A kind of AI dress dimension method for inspecting and system based on YOLO algorithm |
| CN110309870A (en) * | 2019-06-24 | 2019-10-08 | 北京工业大学 | A kind of method of image classification |
-
2020
- 2020-01-17 CN CN202010053985.XA patent/CN111222489A/en not_active Withdrawn
- 2020-06-23 CN CN202010582244.0A patent/CN111507316A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2022217989A1 (en) * | 2021-04-17 | 2022-10-20 | 华为技术有限公司 | Communication port management method and related device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111507316A (en) | 2020-08-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109767422A (en) | Pipe detection recognition methods, storage medium and robot based on deep learning | |
| WO2022017163A1 (en) | Image processing method and apparatus, and device and storage medium | |
| CN110967339B (en) | Method and device for analyzing corn ear characters and corn character analysis equipment | |
| CN111401418A (en) | Employee dressing specification detection method based on improved Faster r-cnn | |
| CN112560816A (en) | Equipment indicator lamp identification method and system based on YOLOv4 | |
| CN112258491A (en) | Three-dimensional simulation method for defects of power equipment | |
| CN110689539A (en) | Workpiece surface defect detection method based on deep learning | |
| CN112215821A (en) | Photovoltaic cell panel grid line defect detection method and system based on artificial intelligence | |
| CN113515655A (en) | Fault identification method and device based on image classification | |
| CN115359239A (en) | Wind power blade defect detection and positioning method and device, storage medium and electronic equipment | |
| CN111507316A (en) | Intelligent identification method and system for optical cable cross connecting box | |
| CN111402250A (en) | Machine vision defect detection method and platform based on edge calculation | |
| CN113205507A (en) | Visual question answering method, system and server | |
| CN111126402B (en) | Image processing method and device, electronic equipment and storage medium | |
| CN111275756A (en) | Spool positioning method and apparatus | |
| CN114330548A (en) | Insulator detection method based on background classification and transfer learning | |
| CN113989684A (en) | Method and system for marking and grading machine inspection defect picture images | |
| CN115294440A (en) | Power transmission line identification method and device and electronic equipment | |
| CN114299307A (en) | Power transmission line image annotation method and related device | |
| CN114757939A (en) | Optical cable production detection supervisory systems based on machine vision | |
| CN113032597A (en) | Power transmission equipment classification method and system based on image processing | |
| CN113947563A (en) | Cable process quality dynamic defect detection method based on deep learning | |
| CN113191295A (en) | Dragon fruit maturity identification method based on image identification | |
| CN112507838A (en) | Pointer meter identification method and device and power inspection robot | |
| CN114897910B (en) | Internal reference and external reference combined optical fiber color mixing identification method, device, equipment and medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20200602 |
|
| WW01 | Invention patent application withdrawn after publication |