CN116343120A - Method for realizing unmanned on duty of radar station based on image recognition of deep learning - Google Patents
Method for realizing unmanned on duty of radar station based on image recognition of deep learning Download PDFInfo
- Publication number
- CN116343120A CN116343120A CN202310311283.0A CN202310311283A CN116343120A CN 116343120 A CN116343120 A CN 116343120A CN 202310311283 A CN202310311283 A CN 202310311283A CN 116343120 A CN116343120 A CN 116343120A
- Authority
- CN
- China
- Prior art keywords
- image data
- recognition
- duty
- text
- radar station
- 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
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/14—Image acquisition
- G06V30/146—Aligning or centring of the image pick-up or image-field
- G06V30/147—Determination of region of interest
-
- 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
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N7/00—Television systems
- H04N7/18—Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/02—Recognising information on displays, dials, clocks
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Signal Processing (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a deep learning-based method for realizing unmanned on duty of a radar station by picture identification, which comprises the following steps of: acquiring image data of a radar station monitoring screen through a camera device and storing the image data into a database; w2: retrieving historical image data and real-time image data in a database, and preprocessing the historical image data and the real-time image data: w3: extracting the shared convolution characteristics of the preprocessed historical image data, and training a character detection and character recognition model to obtain an optimal model of character detection and character recognition; w4: feeding the preprocessed real-time image data to an optimal model to obtain characters of the region of interest; w5: and constructing an alarm dictionary, comparing the characters of the predicted region of interest with the alarm dictionary, and generating abnormal alarm information if the characters of the predicted region of interest appear in the alarm dictionary. The invention effectively reduces the monitoring pressure of monitoring service personnel and improves the efficiency of daily monitoring and abnormal handling work.
Description
Technical Field
The invention relates to the technical field of radar station operation and maintenance management, in particular to a method for realizing unmanned on duty of a radar station based on image recognition of deep learning.
Background
Recently, with the rise of deep learning technology, the development of the field of computer vision has been greatly promoted. The method has extremely outstanding directional contributions including target detection, image segmentation, face detection and recognition, OCR and the like. The image recognition technology is applied to various fields of medicine, military, finance and the like; the image recognition needs clear images of certificates and pure backgrounds, and the robustness and the universality of the recognition method are limited. The problem is thoroughly solved by deep learning, and the deep learning has strong robustness and universality and is not limited by complex background and image quality;
based on the operation and maintenance management of the radar station, the FOTS is a frame integrating detection and identification, and has the characteristics of small model, high speed, high precision, support for multiple angles and the like. The following four types of errors are greatly reduced, and some text areas are omitted; mistakes some non-text regions for text regions; falsely splitting the entire text region into a plurality of separate portions; several independent text regions are falsely merged together.
The invention relates to a method for preprocessing data of stored images, which uses a neural network based on convolution as a feature extraction means. Because of the strong learning capability of CNN, the robustness of feature extraction can be enhanced by matching with a large amount of data, and the robustness can be good when the problems of images such as blurring, distortion, complex background, unclear light and the like are faced.
Disclosure of Invention
The invention aims to provide a deep learning-based image recognition method for realizing unmanned on duty of a radar station, which is used for acquiring image data of a computer screen of a duty room of the radar station based on image acquisition equipment, storing the image data by utilizing a big data correlation technology, acquiring interested fields by adopting a deep learning-based image recognition technology for recognition, and realizing the replacement of a manual field monitoring screen. If the field of interest is abnormal, the quick identification can be realized, and the mobile phone terminal is adopted to inform the relevant responsible person of the alarm information in real time.
The aim of the invention can be achieved by the following technical scheme:
a method for realizing unmanned on duty of a radar station based on image recognition of deep learning comprises the following steps:
w1: image acquisition: acquiring image data of a radar station monitoring screen through a camera device and storing the image data into a database;
w2: retrieving historical image data and real-time image data in a database, and preprocessing the historical image data and the real-time image data:
w3: extracting the shared convolution characteristics of the preprocessed historical image data, and training a character detection and character recognition model to obtain an optimal model of character detection and character recognition;
w4: feeding the preprocessed real-time image data to an optimal model, and predicting to obtain characters of the region of interest;
w5: constructing an alarm dictionary, comparing the characters of the predicted region of interest with the alarm dictionary, and generating abnormal alarm information if the characters of the predicted region of interest appear in the alarm dictionary; otherwise, the processing is not performed, and the process is ended.
As a further scheme of the invention: in W1, the image pickup device photographs the monitor screen once every 2 minutes.
As a further scheme of the invention: in W2, preprocessing of the image data includes blur removal, image enhancement, and light correction.
As a further scheme of the invention: in W3, the characteristic extraction is carried out on the shared convolution characteristic of the historical image data through a CNN neural network.
As a further scheme of the invention: in W3, training the text detection and text recognition model specifically comprises the following steps:
s1: selecting n pieces of image preprocessed historical image data as x1, x2, x3, & gt, xt, xn, and taking the preprocessed historical image data as the input of a training model;
s2: extracting sharing characteristics of character detection and character recognition;
s3: constructing a full convolution network for text detection, and predicting text labels through sharing regional features extracted by convolution and converted by Roirote;
s4: training the network, feeding the preprocessed historical image data for training to obtain an optimal model of character recognition, and using the optimal model for recognition and prediction of the interested text in the real-time image data.
As a further scheme of the invention: and S4, constructing an optimal model by adopting dice loss, and using the dice loss in character detection branches to classify input historical image data into text areas or non-text areas.
As a further scheme of the invention: in S4, the IOU penalty is used in constructing the optimal model for generating the correct bounding box around the text region.
As a further scheme of the invention: and S4, constructing an optimal model by adopting CTC loss, and converting the text in the bounding box for detecting branch prediction into actual text by training text recognition.
As a further scheme of the invention: in W5, the identification information of the real-time image data in which the warning has occurred is stored in the database.
As a further scheme of the invention: and in W5, transmitting the real-time image data with the alarm to a mobile phone terminal of the person on duty for alarm.
The invention has the beneficial effects that: the invention acquires the image data of the computer screen of the radar station duty room based on the image acquisition equipment, stores the image data by utilizing a big data correlation technique, acquires the interested field by adopting a picture identification technique based on deep learning for identification, and realizes the replacement of a manual on-site monitoring screen. If the field of interest is abnormal, the quick identification can be realized, and the mobile phone terminal is adopted to inform the relevant responsible person of the alarm information in real time. The monitoring pressure of monitoring service personnel is effectively reduced, and the efficiency of daily monitoring and abnormal handling work is improved.
Drawings
The invention is further described below with reference to the accompanying drawings.
Fig. 1 is a schematic diagram of the flow chart of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1, the invention discloses a method for realizing unmanned on duty of a radar station based on image recognition of deep learning, which comprises the following steps:
step one: acquiring an image: the camera track is arranged in a duty room of the radar station, the focal length and the angle corresponding to the camera are adjusted, the camera shoots a computer screen every 2 minutes, and the computer screen picture is stored for model training and image recognition;
step two: image preprocessing: the preprocessing process comprises geometric transformation (perspective, distortion and rotation), distortion correction, deblurring, image enhancement and light correction, realizes the processing of stored pictures, and uses a CNN-based neural network;
step three: preprocessing historical image data, constructing an optimal model for training character detection and character recognition, selecting a text region of interest in an image by a frame, and recognizing character content;
unifying text detection and recognition into the same workflow, utilizing an end-to-end framework FOTS (Fast Oriented Text Spotting);
the detection task and the identification task of FOTS share a convolution feature map,
on the one hand, the convolution characteristic is utilized for detection; on the other hand, the method introduces the Roirote for extracting the operator of the directional text region;
after obtaining the text candidate characteristics, inputting the text candidate characteristics into an RNN encoder and a CTC decoder for recognition;
step four: preprocessing the real-time data, and then feeding the real-time data to a stored optimal model to obtain characters of an interested region predicted according to the real-time image data;
step five: and constructing an alarm dictionary, if sensitive words such as abnormality, faults and the like appear in the interested field according to the actual demands of the service, storing the image identification information into a database, and timely transmitting the image identification information to a mobile phone terminal of a relevant person on duty for alarm.
In the third step, the specific steps of training the character detection and character recognition model are as follows:
s1: selecting n pieces of image preprocessing history data as x1, x2, x3, & gt, xt, xn; taking the data after image preprocessing as the input of a training model;
s2: extracting sharing characteristics of character detection and character recognition;
s3: constructing a full convolution network for text detection, and predicting text labels through sharing regional features extracted by convolution and converted by Roirote;
s4: training a network, feeding the constructed preprocessed image:
s41: the dice loss is adopted and used for character detection branches, and input images are classified into text areas or non-text areas;
s42: employing IOU (intersection over union) penalty for generating a correct bounding box around the text region;
s43: CTC (Connectionist Temporal Categical) penalty for training text recognition to convert text in the bounding box that detects branch predictions to actual text;
s5: and storing an optimal character recognition model for recognition prediction of the text of interest in the new image data.
The method has the core points that the image data of the computer screen of the duty room of the radar station is acquired through the image acquisition equipment, the image data is stored by utilizing a big data correlation technique, and the field of interest is acquired for identification by adopting a picture identification technique based on deep learning, so that the screen is monitored on site instead of manual work. If the field of interest is abnormal, the quick identification can be realized, and the mobile phone terminal is adopted to inform the relevant responsible person of the alarm information in real time. The monitoring pressure of monitoring service personnel is effectively reduced, and the efficiency of daily monitoring and abnormal handling work is improved.
The foregoing describes one embodiment of the present invention in detail, but the description is only a preferred embodiment of the present invention and should not be construed as limiting the scope of the invention. All equivalent changes and modifications within the scope of the present invention are intended to be covered by the present invention.
Claims (10)
1. The method for realizing the unmanned on duty of the radar station based on the image recognition of the deep learning is characterized by comprising the following steps:
w1: image acquisition: acquiring image data of a radar station monitoring screen through a camera device and storing the image data into a database;
w2: retrieving historical image data and real-time image data in a database, and preprocessing the historical image data and the real-time image data:
w3: extracting the shared convolution characteristics of the preprocessed historical image data, and training a character detection and character recognition model to obtain an optimal model of character detection and character recognition;
w4: feeding the preprocessed real-time image data to an optimal model, and predicting to obtain characters of the region of interest;
w5: constructing an alarm dictionary, comparing the characters of the predicted region of interest with the alarm dictionary, and generating abnormal alarm information if the characters of the predicted region of interest appear in the alarm dictionary; otherwise, the processing is not performed, and the process is ended.
2. The method for realizing unmanned on duty of a radar station based on picture recognition of deep learning as set forth in claim 1, wherein in W1, the camera device shoots the monitoring screen once every 2 minutes.
3. The method for realizing unmanned on duty of a radar station according to claim 1, wherein in W2, the preprocessing of the image data includes removing blur, image enhancement and light correction.
4. The method for realizing unmanned on duty of a radar station based on image recognition of deep learning as set forth in claim 1, wherein in W3, the feature extraction is performed on the shared convolution feature of the historical image data through a CNN neural network.
5. The method for realizing unmanned on duty of a radar station based on image recognition of deep learning as set forth in claim 1, wherein in W3, training the text detection and text recognition model comprises the following specific steps:
s1: selecting n pieces of image preprocessed historical image data as x1, x2, x3, & gt, xt, xn, and taking the preprocessed historical image data as the input of a training model;
s2: extracting sharing characteristics of character detection and character recognition;
s3: constructing a full convolution network for text detection, and predicting text labels through sharing regional features extracted by convolution and converted by Roirote;
s4: training the network, feeding the preprocessed historical image data for training to obtain an optimal model of character recognition, and using the optimal model for recognition and prediction of the interested text in the real-time image data.
6. The method for realizing unmanned on duty of a radar station based on image recognition of deep learning as set forth in claim 5, wherein in S4, dice loss is used in constructing an optimal model for word detection branches, and input historical image data is classified into text area or non-text area.
7. The method for realizing unmanned on duty of a radar station based on picture recognition of deep learning as set forth in claim 6, wherein in S4, IOU loss is used in constructing an optimal model for generating a correct bounding box around a text region.
8. The method for realizing unmanned on duty of a radar station based on picture recognition of deep learning as set forth in claim 7, wherein in S4 CTC loss is used in constructing an optimal model for training text recognition to convert text in a bounding box of a detected branch prediction into actual text.
9. The method for realizing unmanned on duty of a radar station based on picture recognition of deep learning as set forth in claim 1, wherein in W5, the recognition information of the real-time image data in which the warning occurs is stored in a database.
10. The method for realizing unmanned on duty of a radar station based on image recognition of deep learning as set forth in claim 1, wherein in W5, the real-time image data of the alarm is transmitted to the mobile phone terminal of the person on duty for the alarm.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310311283.0A CN116343120A (en) | 2023-03-27 | 2023-03-27 | Method for realizing unmanned on duty of radar station based on image recognition of deep learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310311283.0A CN116343120A (en) | 2023-03-27 | 2023-03-27 | Method for realizing unmanned on duty of radar station based on image recognition of deep learning |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116343120A true CN116343120A (en) | 2023-06-27 |
Family
ID=86887396
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310311283.0A Pending CN116343120A (en) | 2023-03-27 | 2023-03-27 | Method for realizing unmanned on duty of radar station based on image recognition of deep learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116343120A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117056455A (en) * | 2023-07-04 | 2023-11-14 | 中国经济信息社有限公司 | Manuscript content security auditing method and device, electronic equipment and medium |
-
2023
- 2023-03-27 CN CN202310311283.0A patent/CN116343120A/en active Pending
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117056455A (en) * | 2023-07-04 | 2023-11-14 | 中国经济信息社有限公司 | Manuscript content security auditing method and device, electronic equipment and medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110826538A (en) | Abnormal off-duty identification system for electric power business hall | |
CN110705405A (en) | Target labeling method and device | |
CN112183313B (en) | SlowFast-based power operation field action identification method | |
CN113962274B (en) | Abnormity identification method and device, electronic equipment and storage medium | |
CN111460967A (en) | Illegal building identification method, device, equipment and storage medium | |
CN116343120A (en) | Method for realizing unmanned on duty of radar station based on image recognition of deep learning | |
CN110096945B (en) | Indoor monitoring video key frame real-time extraction method based on machine learning | |
CN110555420A (en) | fusion model network and method based on pedestrian regional feature extraction and re-identification | |
CN112148555A (en) | Intelligent reading and identifying method and system for fault warning information | |
CN112967224A (en) | Electronic circuit board detection system, method and medium based on artificial intelligence | |
CN117593244A (en) | Film product defect detection method based on improved attention mechanism | |
CN110378241B (en) | Crop growth state monitoring method and device, computer equipment and storage medium | |
CN111047731A (en) | AR technology-based telecommunication room inspection method and system | |
CN111738264A (en) | Intelligent acquisition method for data of display panel of machine room equipment | |
CN116030050A (en) | On-line detection and segmentation method for surface defects of fan based on unmanned aerial vehicle and deep learning | |
CN116385465A (en) | Image segmentation model construction and image segmentation method, system, equipment and medium | |
CN114581911B (en) | Steel coil label identification method and system | |
CN115953744A (en) | Vehicle identification tracking method based on deep learning | |
CN116259008A (en) | Water level real-time monitoring method based on computer vision | |
CN114841932A (en) | Foreign matter detection method, system, equipment and medium for photovoltaic panel of photovoltaic power station | |
CN115188049A (en) | Cell face posture detection system and method based on online reinforcement learning | |
CN113935466A (en) | Multitask learning method for intelligent inspection robot of data center | |
Hwang et al. | Deep U-NET based heating film defect inspection system | |
Wu et al. | Express parcel detection based on improved faster regions with CNN features | |
CN114710713B (en) | Automatic video abstract generation method based on deep learning |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication |