CN111598094B - Angle regression instrument reading identification method, equipment and system based on deep learning - Google Patents
Angle regression instrument reading identification method, equipment and system based on deep learning Download PDFInfo
- Publication number
- CN111598094B CN111598094B CN202010461716.7A CN202010461716A CN111598094B CN 111598094 B CN111598094 B CN 111598094B CN 202010461716 A CN202010461716 A CN 202010461716A CN 111598094 B CN111598094 B CN 111598094B
- Authority
- CN
- China
- Prior art keywords
- image
- instrument
- pointer
- training
- meter
- 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.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- 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
- G06F18/217—Validation; Performance evaluation; Active pattern learning techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
-
- 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
- G06V2201/07—Target detection
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Software Systems (AREA)
- General Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Multimedia (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method, equipment and a system for identifying readings of an angle regression instrument based on deep learning, and belongs to the technical field of image identification. According to the invention, the instrument pointer image in the training image is rotated, and a new training image is generated with the background image except the instrument pointer image, and the new training image corresponds to the rotation angle; according to the new training image, the angular regression network is trained, compared with the traditional vision-based identification method, the method not only avoids dependence on the position and light of a camera, but also realizes the identification of the instrument reading through deep learning, and improves the applicability and the identification precision.
Description
Technical Field
The invention relates to the technical field of image recognition, in particular to a method, equipment and a system for recognizing angular regression instrument readings based on deep learning.
Background
In recent years, with the development of deep learning technology and the application in the monitoring field, especially in the image and video monitoring field, the recognition of the meter reading in the image is enabled, so that a meter reading recognition method is needed to realize the recognition of the meter reading in the image.
The prior art provides a meter reading identification method comprising the following steps: the method comprises the steps of performing template matching on an instrument image acquired by a reading camera, performing template matching on a known template and an existing image, comparing each pixel point of an input image with the template, calculating a value for each pixel point, recording the similarity degree of each pixel point after comparison with the template, selecting the point closest to the template, selecting a template corresponding mode, and performing next identification; and reading instrument data, performing simple image processing on the images, entering respective reading channels to perform reading identification, and finally displaying the readings.
The prior art has the following problems:
1. the position requirements for the camera are severe, so that the decrease leads to a decrease in applicability.
2. The image segmentation adopted by the method is based on image gray value segmentation, the dependence on light is relatively large, and once the external environment is darkened or lightened, segmentation failure is easy to cause, so that reading failure is caused.
Disclosure of Invention
In order to solve the problems in the prior art, the invention provides a method, equipment and a system for identifying the reading of an angle regression instrument based on deep learning, which comprise the following steps:
in one aspect, a method for identifying angular regression meter readings based on deep learning is provided, the method comprising:
rotating the instrument pointer image in the training image, and generating a new training image with a background image except the instrument pointer image, wherein the new training image corresponds to the rotation angle;
training an angle regression network according to the new training image;
acquiring an anchor frame where an instrument image in an image to be detected is located;
identifying the angle of the instrument pointer in the anchor frame through the angle regression network;
and mapping the angle of the meter pointer and the meter to obtain meter reading.
Optionally, the rotating the meter pointer image in the training image and generating a new training image with the background image except for the meter pointer image includes:
acquiring an instrument pointer image in the training image;
rotating the meter pointer image according to different rotation angles to obtain meter pointer images corresponding to the rotation angles respectively;
and respectively generating a plurality of new training images by respectively combining the instrument pointer images corresponding to the plurality of rotation angles with the background images.
Optionally, the training the angular regression network according to the new training image includes:
enhancing the new training image to obtain a plurality of enhanced training images;
and training the angle regression network according to the plurality of enhanced training images.
Optionally, the obtaining the anchor frame where the instrument image in the image to be measured is located includes:
and inputting the image to be detected into a target detection algorithm, and outputting an anchor frame where the instrument image is positioned and the instrument type indicated by the instrument image.
Optionally, the mapping the angle of the meter pointer and the meter, and obtaining the meter reading includes:
the instrument readings and the pointer angles are uniformly distributed, and the instrument readings are obtained according to a first mapping formula and the instrument type;
and if the instrument readings and the pointer angles are unevenly distributed, acquiring the instrument readings according to a second mapping formula and the instrument type.
In another aspect, there is provided a deep learning based angle regression meter reading identification device, the device comprising:
the processing module is used for rotating the instrument pointer image in the training image and generating a new training image with the background image except the instrument pointer image, wherein the new training image corresponds to the rotation angle;
the training module is used for training the angular regression network according to the new training image;
the identification module is used for acquiring an anchor frame where the instrument image in the image to be detected is located;
the identification module is also used for identifying the angle of the instrument pointer in the anchor frame through the angle regression network;
the processing module is also used for mapping the angle of the meter pointer and the meter to obtain meter reading.
Optionally, the processing module is specifically configured to:
acquiring an instrument pointer image in the training image;
rotating the meter pointer image according to different rotation angles to obtain meter pointer images corresponding to the rotation angles respectively;
and respectively generating a plurality of new training images by respectively combining the instrument pointer images corresponding to the plurality of rotation angles with the background images.
Optionally, the training module is specifically configured to:
enhancing the new training image to obtain a plurality of enhanced training images;
and training the angle regression network according to the plurality of enhanced training images.
Optionally, the identification module is specifically configured to:
and inputting the image to be detected into a target detection algorithm, and outputting an anchor frame where the instrument image is positioned and the instrument type indicated by the instrument image.
Optionally, the processing module is specifically configured to:
the instrument readings and the pointer angles are uniformly distributed, and the instrument readings are obtained according to a first mapping formula and the instrument type;
and if the instrument readings and the pointer angles are unevenly distributed, acquiring the instrument readings according to a second mapping formula and the instrument type.
In another aspect, a deep learning based angle regression meter reading identification device is provided, the device comprising a processor and a memory coupled to the processor, the memory for storing a set of program code, the processor executing the program code stored by the memory for:
rotating the instrument pointer image in the training image, and generating a new training image with a background image except the instrument pointer image, wherein the new training image corresponds to the rotation angle;
training an angle regression network according to the new training image;
acquiring an anchor frame where an instrument image in an image to be detected is located;
identifying the angle of the instrument pointer in the anchor frame through the angle regression network;
and mapping the angle of the meter pointer and the meter to obtain meter reading.
Optionally, the processor executes the program code stored in the memory, and is specifically configured to perform the following operations:
acquiring an instrument pointer image in the training image;
rotating the meter pointer image according to different rotation angles to obtain meter pointer images corresponding to the rotation angles respectively;
and respectively generating a plurality of new training images by respectively combining the instrument pointer images corresponding to the plurality of rotation angles with the background images.
Optionally, the processor executes the program code stored in the memory, and is specifically configured to perform the following operations:
enhancing the new training image to obtain a plurality of enhanced training images;
and training the angle regression network according to the plurality of enhanced training images.
Optionally, the processor executes the program code stored in the memory, and is specifically configured to perform the following operations:
and inputting the image to be detected into a target detection algorithm, and outputting an anchor frame where the instrument image is positioned and the instrument type indicated by the instrument image.
Optionally, the processor executes the program code stored in the memory, and is specifically configured to perform the following operations:
the instrument readings and the pointer angles are uniformly distributed, and the instrument readings are obtained according to a first mapping formula and the instrument type;
and if the instrument readings and the pointer angles are unevenly distributed, acquiring the instrument readings according to a second mapping formula and the instrument type.
In another aspect, there is provided a deep learning based angle regression meter reading identification system, the system comprising:
the processing equipment is used for rotating the instrument pointer image in the training image and generating a new training image with the background image except the instrument pointer image, wherein the new training image corresponds to the rotation angle;
training equipment for training the angular regression network according to the new training image;
the identification equipment is used for acquiring an anchor frame where the instrument image in the image to be detected is located;
the identification equipment is also used for identifying the angle of the instrument pointer in the anchor frame through the angle regression network;
the processing device is also used for mapping the angle of the meter pointer and the meter to obtain meter readings.
Optionally, the processing device is specifically configured to:
acquiring an instrument pointer image in the training image;
rotating the meter pointer image according to different rotation angles to obtain meter pointer images corresponding to the rotation angles respectively;
and respectively generating a plurality of new training images by respectively combining the instrument pointer images corresponding to the plurality of rotation angles with the background images.
Optionally, the training device is specifically configured to:
enhancing the new training image to obtain a plurality of enhanced training images;
and training the angle regression network according to the plurality of enhanced training images.
Optionally, the identifying device is specifically configured to:
and inputting the image to be detected into a target detection algorithm, and outputting an anchor frame where the instrument image is positioned and the instrument type indicated by the instrument image.
Optionally, the processing device is specifically configured to:
the instrument readings and the pointer angles are uniformly distributed, and the instrument readings are obtained according to a first mapping formula and the instrument type;
and if the instrument readings and the pointer angles are unevenly distributed, acquiring the instrument readings according to a second mapping formula and the instrument type.
The invention provides a method, equipment and a system for identifying readings of an angle regression instrument based on deep learning, which comprise the following steps: rotating the meter pointer image in the training image, and generating a new training image with the background image except the meter pointer image, wherein the new training image corresponds to the rotation angle; training the angle regression network according to the new training image; acquiring an anchor frame where an instrument image in an image to be detected is located; identifying the angle of the instrument pointer in the anchor frame through an angle regression network; and mapping the angle of the meter pointer and the meter to obtain meter reading.
The technical scheme provided by the invention has the beneficial effects that:
generating a new training image by rotating the meter pointer image in the training image and generating a background image except the meter pointer image, wherein the new training image corresponds to the rotation angle; according to the new training image, the angular regression network is trained, compared with the traditional vision-based identification method, the method not only avoids dependence on the position and light of a camera, but also realizes the identification of the instrument reading through deep learning, and improves the applicability and the identification precision.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, and it is apparent that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of an angle regression meter reading identification method based on deep learning provided by an embodiment of the invention;
FIG. 2 is a flow chart of a method for identifying angular regression meter readings based on deep learning according to an embodiment of the invention;
FIG. 3 is a schematic diagram of an angle regression instrument reading identification device based on deep learning according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of an angle regression instrument reading identification device based on deep learning according to an embodiment of the present invention;
fig. 5 is a schematic diagram of an angle regression meter reading identification system based on deep learning according to an embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings.
For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments of the present invention. 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.
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. 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.
Embodiments of the invention are described in further detail below with reference to the drawings.
Example 1
The embodiment of the invention provides a method for identifying the readings of an angle regression instrument based on deep learning, which is shown by referring to FIG. 1 and comprises the following steps:
101. and rotating the meter pointer image in the training image, and generating a new training image with the background image except the meter pointer image, wherein the new training image corresponds to the rotation angle.
Specifically, an instrument pointer image in a training image is obtained;
rotating the meter pointer image according to different rotation angles to obtain meter pointer images corresponding to the rotation angles respectively;
and respectively generating a plurality of new training images by respectively combining the instrument pointer images corresponding to the rotation angles with the background images.
102. And training the angle regression network according to the new training image.
Specifically, a new training image is enhanced, and a plurality of enhanced training images are obtained;
and training the angle regression network according to the plurality of enhanced training images.
103. And acquiring an anchor frame where the instrument image in the image to be detected is located.
Specifically, the image to be detected is input into a target detection network, and an anchor frame where the instrument image is located and the instrument type indicated by the instrument image are output.
104. And identifying the angle of the instrument pointer in the anchor frame through an angle regression network.
105. And mapping the angle of the meter pointer and the meter to obtain meter reading.
Specifically, if the instrument readings and the pointer angles are uniformly distributed, acquiring the instrument readings according to a first mapping formula and the instrument type;
and if the instrument readings and the pointer angles are unevenly distributed, acquiring the instrument readings according to a second mapping formula and the instrument type.
Example two
The embodiment of the invention provides a method for identifying the readings of an angle regression instrument based on deep learning, which is shown by referring to FIG. 2 and comprises the following steps:
the image according to the embodiment of the invention can be taken by a fixed device, which can be a monitoring camera or an imaging device for recording, such as a vehicle-mounted recorder, or can be a mobile device, which can be a mobile phone or a special imaging device, such as a camera or a handheld recorder.
201. And acquiring an instrument pointer image in the training image.
Specifically, the acquisition process can be realized through an image recognition algorithm, or can be completed through manual setting by a user; the manual setting process of the user can be as follows:
acquiring a geometric area which at least comprises the pointer image of the instrument and is input by a gesture or a keyboard on a display screen of a user;
identifying a meter pointer image in the geometric region; or alternatively
Acquiring the outline of the pointer image of the instrument, which is input by gestures on a display screen of a user;
identifying a meter pointer image in the geometric region;
wherein the display screen comprises at least the training image.
202. And rotating the meter pointer image according to different rotation angles to obtain meter pointer images corresponding to the rotation angles respectively.
Specifically, the rotation process can be realized through a preset algorithm, and the instrument pointer is input into the preset algorithm to obtain instrument pointer images corresponding to a plurality of different rotation angles respectively; the rotation angle may be any angle within the meter reading range.
In order to further increase the number of training images and improve the accuracy of instrument reading identification, the reading range can be divided according to the minimum reading of the instrument reading, and a plurality of rotation angles can be obtained.
203. And respectively generating a plurality of new training images by respectively combining the instrument pointer images corresponding to the rotation angles with the background images.
Specifically, the process may be implemented by an image synthesis algorithm, and the embodiment of the present invention does not limit the specific synthesis algorithm.
It should be noted that, steps 201 to 203 are processes of rotating the meter pointer image in the training image and generating a new training image with the background image except for the meter pointer image, and the processes may be implemented in other ways besides the processes described in the above steps, which are not limited in the embodiment of the present invention.
In addition, the above process is implemented on one training image, and in practical application, in order to further improve accuracy, the above process may be performed on a plurality of training images to increase the sample size.
204. And enhancing the new training image to obtain a plurality of enhanced training images.
Specifically, at least one of the following steps is performed for any one of a plurality of new training images;
noise adjustment is carried out on the new training image, and a training image after noise adjustment is obtained; the adjustment mode comprises increasing or decreasing the image noise parameter; or alternatively
Brightness adjustment is carried out on the new training image, and a training image with brightness adjusted is obtained; the adjustment mode comprises increasing or decreasing the image brightness parameter; or alternatively
Performing perspective transformation on the new training image; obtaining a training image after perspective transformation;
and continuing to perform at least one of the above steps on the other new training images until the above steps are completed on all the new training images.
By enhancing the new training images, the number of the training images is increased, so that the recognition accuracy of the angle regression algorithm after training is improved, and the accuracy of instrument reading recognition is further improved.
205. And training the angle regression network according to the plurality of enhanced training images.
Specifically, the angle regression network may be: a target recognition network with input data as images and output data as numerical values;
the training process is as follows:
setting training parameters;
acquiring anchor frames where the instrument images are located in the plurality of enhanced training images;
inputting an anchor frame where the instrument image is located into an angle regression network;
training the angle regression network according to the training parameters, and obtaining a training result; the training result at least comprises a mapping relation between the pointer angle and the position before reading;
if the training result does not meet the expectation, resetting the training parameters or adjusting the training times until the training result meets the expectation.
It should be noted that, steps 204 to 205 are implemented in a process of training the angular regression network according to the new training image, and the process may be implemented in other ways besides the process described in the above steps, which is not limited to the specific way in the embodiment of the present invention.
Steps 201 to 205 are the process of training the angular regression network. Steps 206 through 209 are processes for identifying meter readings in the image to be measured using the angle regression network.
206. And acquiring an anchor frame where the instrument image in the image to be detected is located.
Specifically, the labeling category is set as the instrument type;
inputting an image to be detected into a target detection network, outputting an anchor frame where an instrument image is positioned and an instrument type; the object detection network may be a network of the YOLOv3 algorithm.
In addition to the above-described steps, the process may be implemented in other manners, and embodiments of the present invention are not limited to specific manners.
Because the accuracy of the YOLOv3 algorithm for identifying the small target image is high, the accuracy of the anchor frame where the instrument image is located in the image to be detected, namely the instrument image, is obtained through the YOLOv3 algorithm, and therefore the accuracy of the reading identification of the instrument is improved.
207. And identifying the angle of the instrument pointer in the anchor frame through an angle regression network.
Specifically, the anchor frame is input into the angle regression network, and the angle of the instrument pointer in the anchor frame is output.
208. And if the meter readings and the pointers are uniformly distributed, acquiring the meter readings according to the first mapping formula and the meter type.
Specifically, if the meter reading and the pointer are uniformly distributed on the dial of the meter indicated by the meter type, the first mapping formula is:
y=ax+b；
where x is the angle, y is the reading, and a and b are preset parameters related to the meter indicated by the meter type.
209. And if the meter readings and the pointers are unevenly distributed, acquiring the meter readings according to a second mapping formula and the meter type.
Specifically, if the meter reading and the pointer are unevenly distributed on the dial of the meter indicated by the meter type, the second mapping formula is:
y = ax+b，( 0< x < A)；
y = cx+d，(x > A)；
where x is the angle, y is the reading, a, b, c, d and A are preset parameters related to the meter indicated by the meter type.
It should be noted that, step 208 or step 209 is a process of mapping the angle of the meter pointer and the meter, and obtaining the meter reading, and the process may be implemented in other manners besides the process described in the foregoing step, which is not limited to the specific manner in the embodiment of the present invention.
Example III
The embodiment of the invention provides an angle regression instrument reading identification device 3 based on deep learning, which is shown by referring to FIG. 3 and comprises:
the processing module 31 is configured to rotate the meter pointer image in the training image, and generate a new training image with a background image other than the meter pointer image, where the new training image corresponds to the rotation angle;
a training module 32 for training the angular regression network according to the new training image;
the identification module 33 is used for acquiring an anchor frame where the instrument image in the image to be detected is located;
the identification module 33 is further used for identifying the angle of the instrument pointer in the anchor frame through an angle regression network;
the processing module is also used for mapping the angle of the meter pointer and the meter to obtain meter readings.
Optionally, the processing module 31 is specifically configured to:
acquiring an instrument pointer image in a training image;
rotating the meter pointer image according to different rotation angles to obtain meter pointer images corresponding to the rotation angles respectively;
and respectively generating a plurality of new training images by respectively combining the instrument pointer images corresponding to the rotation angles with the background images.
Optionally, the training module 32 is specifically configured to:
enhancing the new training image to obtain a plurality of enhanced training images;
and training the angle regression network according to the plurality of enhanced training images.
Optionally, the identification module 33 is specifically configured to:
and inputting the image to be detected into a target detection algorithm, and outputting an anchor frame where the instrument image is positioned and the instrument type indicated by the instrument image.
Optionally, the processing module 31 is specifically configured to:
the instrument readings and the pointer angles are uniformly distributed, and according to a first mapping formula and the instrument type, the instrument readings are obtained;
and acquiring the meter readings according to the second mapping formula and the meter type if the plurality of meter readings and the pointers are unevenly distributed.
Example IV
An embodiment of the present invention provides an angle regression meter reading identification device 4 based on deep learning, referring to fig. 4, the device includes a processor 41 and a memory 42 connected to the processor 41, the memory 42 is used for storing a set of program codes, and the processor 41 executes the program codes stored in the memory 42 to perform the following operations:
rotating the meter pointer image in the training image, and generating a new training image with the background image except the meter pointer image, wherein the new training image corresponds to the rotation angle;
training the angle regression network according to the new training image;
acquiring an anchor frame where an instrument image in an image to be detected is located;
identifying the angle of the instrument pointer in the anchor frame through an angle regression network;
and mapping the angle of the meter pointer and the meter to obtain meter reading.
Optionally, the processor 41 executes the program code stored in the memory 42 for performing the following operations:
acquiring an instrument pointer image in a training image;
rotating the meter pointer image according to different rotation angles to obtain meter pointer images corresponding to the rotation angles respectively;
and respectively generating a plurality of new training images by respectively combining the instrument pointer images corresponding to the rotation angles with the background images.
Optionally, the processor 41 executes the program code stored in the memory 42 for performing the following operations:
enhancing the new training image to obtain a plurality of enhanced training images;
and training the angle regression network according to the plurality of enhanced training images.
Optionally, the processor 41 executes the program code stored in the memory 42 for performing the following operations:
and inputting the image to be detected into a target detection algorithm, and outputting an anchor frame where the instrument image is positioned and the instrument type indicated by the instrument image.
Optionally, the processor 41 executes the program code stored in the memory 42 for performing the following operations:
the instrument readings and the pointer angles are uniformly distributed, and according to a first mapping formula and the instrument type, the instrument readings are obtained;
and acquiring the meter readings according to the second mapping formula and the meter type if the plurality of meter readings and the pointers are unevenly distributed.
Example five
The embodiment of the invention provides an angle regression instrument reading identification system based on deep learning, which is shown by referring to FIG. 5 and comprises the following components:
a processing device 51 for rotating the meter pointer image in the training image and generating a new training image corresponding to the rotation angle with the background image other than the meter pointer image;
training means 52 for training the angular regression network based on the new training image;
the identifying device 53 is configured to obtain an anchor frame where the instrument image in the image to be detected is located;
the identifying device 53 is further configured to identify an angle of the meter pointer in the anchor frame through an angle regression network;
the processing device 51 is also used to map the angle of the meter pointer to the meter and to take meter readings.
Optionally, the processing device 51 is specifically configured to:
acquiring an instrument pointer image in a training image;
rotating the meter pointer image according to different rotation angles to obtain meter pointer images corresponding to the rotation angles respectively;
and respectively generating a plurality of new training images by respectively combining the instrument pointer images corresponding to the rotation angles with the background images.
Optionally, the training device 52 is specifically configured to:
enhancing the new training image to obtain a plurality of enhanced training images;
and training the angle regression network according to the plurality of enhanced training images.
Optionally, the identifying device 53 is specifically configured to:
and inputting the image to be detected into a target detection algorithm, and outputting an anchor frame where the instrument image is positioned and the instrument type indicated by the instrument image.
Optionally, the processing device 51 is specifically configured to:
the instrument readings and the pointer angles are uniformly distributed, and according to a first mapping formula and the instrument type, the instrument readings are obtained;
and acquiring the meter readings according to the second mapping formula and the meter type if the plurality of meter readings and the pointers are unevenly distributed.
The invention provides a method, equipment and a system for identifying an angle regression instrument reading based on deep learning, which are characterized in that an instrument pointer image in a training image is rotated, a new training image is generated with a background image except the instrument pointer image, and the new training image corresponds to a rotation angle; according to the new training image, the angular regression network is trained, compared with the traditional vision-based identification method, the method not only avoids dependence on the position and light of a camera, but also realizes the identification of the instrument reading through deep learning, and improves the applicability and the identification precision.
Any combination of the above optional solutions may be adopted to form an optional embodiment of the present invention, which is not described herein.
It should be noted that: the meter reading identification device and the system provided in the foregoing embodiments are only exemplified by the division of the foregoing functional modules when the meter reading identification method is executed, and in practical applications, the foregoing functional allocation may be performed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules, so as to complete all or part of the functions described above. In addition, the embodiments of the method, the device and the system for identifying meter readings provided in the foregoing embodiments belong to the same concept, and detailed implementation processes of the embodiments of the method are detailed in the embodiments of the method, which are not repeated herein.
It will be understood by those skilled in the art that all or part of the steps for implementing the above embodiments may be implemented by hardware, or may be implemented by a program for instructing relevant hardware, where the program may be stored in a computer readable storage medium, and the storage medium may be a read-only memory, a magnetic disk or an optical disk, etc.
The foregoing description of the preferred embodiments of the invention is not intended to limit the invention to the precise form disclosed, and any such modifications, equivalents, and alternatives falling within the spirit and scope of the invention are intended to be included within the scope of the invention.
Claims (8)
1. An angle regression instrument reading identification method based on deep learning, which is characterized by comprising the following steps:
rotating the instrument pointer image in the training image, and generating a new training image with a background image except the instrument pointer image, wherein the new training image corresponds to the rotation angle;
training an angle regression network according to the new training image;
acquiring an anchor frame where an instrument image in an image to be detected is located;
identifying the angle of the instrument pointer in the anchor frame through the angle regression network;
mapping the angle of the meter pointer and a meter to obtain meter reading;
wherein generating a new training image comprises:
acquiring an instrument pointer image in the training image;
rotating the meter pointer image according to different rotation angles to obtain meter pointer images corresponding to the rotation angles respectively;
respectively generating a plurality of new training images by respectively combining the instrument pointer images corresponding to the plurality of rotation angles with the background images;
wherein, obtaining the meter pointer image includes:
acquiring a geometric area which at least comprises the pointer image of the instrument and is input by a gesture or a keyboard on a display screen of a user;
a meter pointer image in the geometric region is identified.
2. The method of claim 1, wherein training the angular regression network from the new training image comprises:
enhancing the new training image to obtain a plurality of enhanced training images;
and training the angle regression network according to the plurality of enhanced training images.
3. The method of claim 2, wherein the acquiring the anchor frame in which the meter image in the image to be measured is located comprises:
and inputting the image to be detected into a target detection network, and outputting an anchor frame where the instrument image is positioned and the instrument type indicated by the instrument image.
4. The method of claim 3, wherein mapping the angle of the meter pointer and the meter, obtaining a meter reading comprises: the instrument readings and the pointer angles are uniformly distributed, and the instrument readings are obtained according to a first mapping formula and the instrument type;
and if the instrument readings and the pointer angles are unevenly distributed, acquiring the instrument readings according to a second mapping formula and the instrument type.
5. An angle regression meter reading identification device based on deep learning, applying the method of any of claims 1-4, characterized in that the device comprises:
the processing module is used for rotating the instrument pointer image in the training image and generating a new training image with the background image except the instrument pointer image, wherein the new training image corresponds to the rotation angle;
the training module is used for training the angular regression network according to the new training image;
the identification module is used for acquiring an anchor frame where the instrument image in the image to be detected is located;
the identification module is also used for identifying the angle of the instrument pointer in the anchor frame through the angle regression network;
the processing module is also used for mapping the angle of the meter pointer and the meter to obtain meter reading.
6. The apparatus of claim 5, wherein the processing module is specifically configured to:
acquiring an instrument pointer image in the training image;
rotating the meter pointer image according to different rotation angles to obtain meter pointer images with a plurality of rotation angles;
and generating a plurality of new training images by respectively combining the meter pointer images corresponding to the rotation angles with background images except the meter pointer images.
7. The device according to claim 6, wherein the training module is specifically configured to:
enhancing the new training image to obtain a plurality of enhanced training images;
and training the angle regression network according to the plurality of enhanced training images.
8. The device according to claim 7, wherein the identification module is specifically configured to:
and inputting the image to be detected into a target detection network, and outputting an anchor frame where the instrument image is positioned and the instrument type indicated by the instrument image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010461716.7A CN111598094B (en) | 2020-05-27 | 2020-05-27 | Angle regression instrument reading identification method, equipment and system based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010461716.7A CN111598094B (en) | 2020-05-27 | 2020-05-27 | Angle regression instrument reading identification method, equipment and system based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111598094A CN111598094A (en) | 2020-08-28 |
| CN111598094B true CN111598094B (en) | 2023-08-18 |
Family
ID=72188870
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010461716.7A Active CN111598094B (en) | 2020-05-27 | 2020-05-27 | Angle regression instrument reading identification method, equipment and system based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111598094B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2022148006A (en) * | 2021-03-24 | 2022-10-06 | 株式会社東芝 | Training device, processing system, training method, processing method, program, and storage medium |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106951900A (en) * | 2017-04-13 | 2017-07-14 | 杭州申昊科技股份有限公司 | A kind of automatic identifying method of arrester meter reading |
| CN108764257A (en) * | 2018-05-23 | 2018-11-06 | 郑州金惠计算机系统工程有限公司 | A kind of pointer instrument recognition methods of various visual angles |
| CN109543682A (en) * | 2018-11-23 | 2019-03-29 | 电子科技大学 | A kind of readings of pointer type meters method based on deep learning |
| CN110232376A (en) * | 2019-06-11 | 2019-09-13 | 重庆邮电大学 | A kind of gear type digital instrument recognition methods returned using projection |
| CN110245597A (en) * | 2019-06-06 | 2019-09-17 | 重庆邮电大学 | A kind of pointer instrument versatility recognition methods |
| CN110443242A (en) * | 2019-07-31 | 2019-11-12 | 新华三大数据技术有限公司 | Read frame detection method, Model of Target Recognition training method and relevant apparatus |
| CN110633679A (en) * | 2019-09-19 | 2019-12-31 | 湘潭大学 | Automatic pointer instrument indicating identification method and system based on genetic algorithm |
| CN111046881A (en) * | 2019-12-02 | 2020-04-21 | 许昌北邮万联网络技术有限公司 | Pointer type instrument reading identification method based on computer vision and deep learning |
| CN111104942A (en) * | 2019-12-09 | 2020-05-05 | 熵智科技(深圳)有限公司 | Template matching network training method, template matching network recognition method and template matching network recognition device |
-
2020
- 2020-05-27 CN CN202010461716.7A patent/CN111598094B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106951900A (en) * | 2017-04-13 | 2017-07-14 | 杭州申昊科技股份有限公司 | A kind of automatic identifying method of arrester meter reading |
| CN108764257A (en) * | 2018-05-23 | 2018-11-06 | 郑州金惠计算机系统工程有限公司 | A kind of pointer instrument recognition methods of various visual angles |
| CN109543682A (en) * | 2018-11-23 | 2019-03-29 | 电子科技大学 | A kind of readings of pointer type meters method based on deep learning |
| CN110245597A (en) * | 2019-06-06 | 2019-09-17 | 重庆邮电大学 | A kind of pointer instrument versatility recognition methods |
| CN110232376A (en) * | 2019-06-11 | 2019-09-13 | 重庆邮电大学 | A kind of gear type digital instrument recognition methods returned using projection |
| CN110443242A (en) * | 2019-07-31 | 2019-11-12 | 新华三大数据技术有限公司 | Read frame detection method, Model of Target Recognition training method and relevant apparatus |
| CN110633679A (en) * | 2019-09-19 | 2019-12-31 | 湘潭大学 | Automatic pointer instrument indicating identification method and system based on genetic algorithm |
| CN111046881A (en) * | 2019-12-02 | 2020-04-21 | 许昌北邮万联网络技术有限公司 | Pointer type instrument reading identification method based on computer vision and deep learning |
| CN111104942A (en) * | 2019-12-09 | 2020-05-05 | 熵智科技(深圳)有限公司 | Template matching network training method, template matching network recognition method and template matching network recognition device |
Non-Patent Citations (1)
| Title |
|---|
| 基于深度学习的指针式仪表检测与识别研究;徐发兵;吴怀宇;陈志环;喻汉;;高技术通讯(第12期);第46-55页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111598094A (en) | 2020-08-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111539412B (en) | Image analysis method, system, device and medium based on OCR | |
| CN113780201B (en) | Hand image processing method and device, equipment and medium | |
| US10679094B2 (en) | Automatic ruler detection | |
| CN109947273B (en) | Point reading positioning method and device | |
| CN111290684B (en) | Image display method, image display device and terminal equipment | |
| CN110910445B (en) | Object size detection method, device, detection equipment and storage medium | |
| CN115311618A (en) | Assembly quality inspection method based on deep learning and object matching | |
| CN112559341A (en) | Picture testing method, device, equipment and storage medium | |
| CN111598094B (en) | Angle regression instrument reading identification method, equipment and system based on deep learning | |
| CN117372286B (en) | Python-based image noise optimization method and system | |
| CN112613425B (en) | Target identification system for small sample underwater image | |
| CN111815748A (en) | Animation processing method and device, storage medium and electronic equipment | |
| US20230401670A1 (en) | Multi-scale autoencoder generation method, electronic device and readable storage medium | |
| CN111281355A (en) | Method and equipment for determining pulse acquisition position | |
| CN112464753B (en) | Method and device for detecting key points in image and terminal equipment | |
| CN115564828A (en) | Machine vision positioning method and system based on ambiguity algorithm | |
| CN114663681A (en) | Method for reading pointer type meter and related product | |
| CN117372937B (en) | Data reading method based on pointer instrument | |
| JP2014232373A (en) | Collation object extraction system, collation object extraction method, and collation object extraction program | |
| CN107341799A (en) | Diagnostic data acquisition methods and device | |
| CN112287772B (en) | Fingerprint trace detection method, fingerprint detection device and computer readable storage medium | |
| CN117132667B (en) | Thermal image processing method and related device based on environmental temperature feedback | |
| CN112052729B (en) | Intelligent dynamic high-definition video detection method and system based on face recognition | |
| CN118466149A (en) | Clock tester calibration method, device and system | |
| Zhang et al. | Projection segmentation-based image recognition technology for automatic reading of gas meter |
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 |