CN114155432A - Meter reading identification method based on robot - Google Patents
Meter reading identification method based on robot Download PDFInfo
- Publication number
- CN114155432A CN114155432A CN202111309307.6A CN202111309307A CN114155432A CN 114155432 A CN114155432 A CN 114155432A CN 202111309307 A CN202111309307 A CN 202111309307A CN 114155432 A CN114155432 A CN 114155432A
- Authority
- CN
- China
- Prior art keywords
- meter
- robot
- meter reading
- image data
- identification method
- 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
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
Abstract
The invention discloses a meter reading identification method based on a robot, which comprises the steps of establishing a meter reading identification model through deep learning, automatically calling the meter reading identification model through real-time inspection of the robot, accurately identifying the type of a meter and calculating the current reading of the meter; the reading identification of multi-point, multi-scene and multi-type meters is realized; the possible careless omission of manual meter reading is reduced, the labor intensity is effectively reduced, and the operation and maintenance cost of a power plant is reduced; and updating sample image data according to the abnormal check, and improving the robustness of the algorithm.
Description
[ technical field ] A method for producing a semiconductor device
The application relates to the technical field of computers, in particular to a meter reading identification method based on a robot.
[ background of the invention ]
In an industrial production scenario, a large number of gauges exist, such as thermometers, pressure gauges, liquid level gauges, etc., due to their wide application. With the progress of digital application, the digital meter reading of the meter is required to be realized, most of the existing automatic inspection of the meter realizes the timing reading of the meter by additionally arranging a customized camera, and the technical scheme has the defects of large equipment installation workload, fussy maintenance and weak point location expansibility.
[ summary of the invention ]
The invention aims to provide a meter reading identification method based on a robot, which realizes multi-point, multi-scene and multi-type meter reading identification.
The invention provides a meter reading identification method based on a robot, which comprises the following steps:
s1, collecting sample image data of the meter by the robot;
s2, cleaning and labeling the sample image data;
s3, training sample image data through a deep learning algorithm to obtain a meter reading recognition model;
s4, the robot patrols and collects the patrolling image data of the meter;
and S5, automatically calling a meter reading recognition model, and calculating to obtain the current reading of the meter.
Further, the step S1 includes the following steps:
s11, point location configuration is carried out on the meters needing automatic inspection in the factory area through the robot inspection configuration tool;
s12, the robot identifies and adjusts the pose of the robot and/or adjusts the direction of the camera according to the image, so that the center of the camera corresponds to the center of the meter;
and S13, the robot shoots at least one meter image and records corresponding robot position and posture data during shooting.
Further, the labeling in the step S2 includes labeling the meter category.
Further, the labeling in step S2 includes labeling a maximum meter range and a minimum meter range.
Further, the cleaning of the sample image data in step S2 includes performing scaling, rotation, displacement, ray transformation, and image synthesis on the sample image data.
Further, in the step S3, when performing the deep learning, the most accurate image is selected from the plurality of meter images as the patrol sample image.
Further, the meter reading identification model in the step S3 includes a meter type automatic identification model for identifying a meter type; when the robot patrols and examines, firstly calling a meter type automatic identification model to identify the meter type, then photographing the meter, and collecting image data;
further, the meter reading identification model in step S5 detects a straight line by hough transform, extracts a pointer image, fits the angles between the pointer and the maximum measurement range and the minimum measurement range of the meter by the least square method, and calculates the current meter reading by combining the identified maximum measurement range and the minimum measurement range of the meter.
Furthermore, before the straight line is detected by adopting Hough transform, the position of the cloud deck of the robot is calibrated according to the image characteristics of the sample of the inspection point, so that the acquisition positions of the image of the inspection point and the image of the sample of the inspection point are consistent.
Further, still include:
s6, feeding the current meter reading back to the service management background, and checking the abnormal meter reading by the warning early warning module; the abnormal reading of the meter indicates that the current reading of the meter is not in the range allowed by the meter;
and S7, when the meter reading is abnormal, carrying out manual verification, when the actual result of the meter is inconsistent with the result fed back by the meter reading recognition model, marking the image data, adding the sample image data, and repeating the steps S2 to S3 to train and update the meter reading recognition model.
Compared with the prior art, the method has the following advantages:
establishing a meter reading identification model through deep learning, automatically calling the meter reading identification model through real-time inspection of a robot, accurately identifying the type of the meter, and calculating the current reading of the meter; the reading identification of multi-point, multi-scene and multi-type meters is realized; the possible careless omission of manual meter reading is reduced, the labor intensity is effectively reduced, and the operation and maintenance cost of a power plant is reduced; and updating sample image data according to the abnormal check, and improving the robustness of the algorithm.
[ description of the drawings ]
Fig. 1 is a flow chart of a meter reading identification method based on a robot according to the present invention.
[ detailed description ] embodiments
In order to make the aforementioned features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in detail below, but the present invention is not limited thereto.
As shown in fig. 1, a meter reading identification method based on a robot includes the following steps:
s1, collecting sample image data of the meter by the robot; specifically, the method comprises the following steps of,
s11, point location configuration is carried out on the meters needing automatic inspection in the factory area through the robot inspection configuration tool; configuring the specific position of the meter;
s12, the robot identifies and adjusts the pose of the robot and/or adjusts the direction of the camera according to the image, so that the center of the camera corresponds to the center of the meter; the center point of the meter is calculated by identifying the periphery of a meter panel and the pointer, and the pose of a robot and/or the direction of a camera are/is adjusted, so that the axis of the camera is superposed with the center of the meter, and the pointer of a shot picture is ensured to be not inclined; the direction of the robot shooting meter is very important, and if the direction is angular deviation, the reading error is large.
S13, the robot shoots at least one meter image and records corresponding robot position and posture data during shooting; the robot can shoot a plurality of meter images, simultaneously record corresponding robot pose data during shooting, select the most accurate image from the plurality of meter images as an inspection sample image during deep learning, and adopt the robot pose corresponding to the inspection sample image when the robot inspects and collects the image data of the meters;
s2, cleaning and labeling the sample image data; marking the meter type, the maximum meter range and the minimum meter range; carrying out scale transformation, rotation transformation, displacement transformation, light transformation and image synthesis on sample image data; the image size, namely the length and the width, can be changed according to the proportion, namely the length-width ratio, or the image can be enlarged or reduced according to the requirement without according to the length-width ratio; the rotation conversion converts all points on the image into the same direction around a fixed point and rotates by the same angle; the displacement transformation adds the specified horizontal offset and vertical offset to all pixel coordinates of the image respectively; the brightness of the image is adjusted through light ray transformation; image synthesis is to fuse images of angles; finally, the image with unified standard is obtained.
S3, training sample image data through a deep learning algorithm to obtain a meter reading recognition model; the meter reading identification model comprises a meter type automatic identification model for identifying the meter type, and the meter type automatic identification model obtains the characteristics of each meter through a characteristic extraction and classification algorithm and can quickly identify the meter type; the meter reading identification model carries out image calibration on the inspection image through the inspection point sample image, adopts Hough transformation to detect a straight line, extracts a pointer image, fits the angles between the pointer and the maximum measuring range and the minimum measuring range of the meter through a least square method, and calculates the current meter reading by combining the identified maximum measuring range and the minimum measuring range of the meter.
S4, the robot patrols and collects the image data of the meter; when the robot patrols and examines, firstly calling a meter type automatic identification model to identify the meter type, quickly positioning a meter area, quickly finding a meter center, photographing the front of the meter, and collecting image data; removing image interference information; the image data is ensured to be cleaned, complete and free of deflection; and calibrating the position of the robot holder according to the image characteristics of the inspection sample book, so that the positions of the collected image and the image of the inspection sample book are kept consistent.
And S5, automatically calling a meter reading identification model, carrying out image calibration on the inspection image through the inspection point sample image by the meter reading identification model, detecting a straight line by adopting Hough transformation, extracting a pointer image, fitting the angles of the pointer and the maximum measuring range and the minimum measuring range of the meter through a least square method, and calculating the current reading of the meter by combining the identified maximum measuring range and the minimum measuring range of the meter.
S6, feeding back the calculated current meter reading to a service management background, and checking the meter reading abnormity by an alarm early warning module; the abnormal meter reading means that the current meter reading is not within the allowable range of the meter.
And S7, when the meter reading is abnormal, carrying out manual verification, when the actual result of the meter is inconsistent with the result fed back by the meter reading recognition model, marking the image data, adding the sample image data, and repeating the steps S2 to S3 to train and update the meter reading recognition model.
Another embodiment of the method
S1, collecting sample image data of the meter by the robot; specifically, the method comprises the following steps of,
s11, point location configuration is carried out on the meters needing automatic inspection in the factory area through the robot inspection configuration tool, and meanwhile, the maximum measurement range and the minimum measurement range of the meters are configured;
s12, the robot identifies and adjusts the pose of the robot and/or adjusts the direction of the camera according to the image, so that the center of the camera corresponds to the center of the meter;
s13, the robot shoots at least one meter image and records corresponding robot position and posture data during shooting;
s2, cleaning and labeling the sample image data; marking the meter type; .
S3, training sample image data through a deep learning algorithm to obtain a meter reading recognition model;
s4, the robot patrols and collects the image data of the meter;
and S5, automatically calling a meter reading identification model, carrying out image calibration on the inspection image through the inspection point sample image by the meter reading identification model, detecting a straight line by adopting Hough transformation, extracting a pointer image, fitting the angles of the pointer and the maximum measuring range and the minimum measuring range of the meter through a least square method, and calculating the current reading of the meter by combining the identified maximum measuring range and the minimum measuring range of the meter.
S6, feeding back the calculated current meter reading to a service management background, and checking the meter reading abnormity by an alarm early warning module; the abnormal meter reading means that the current meter reading is not within the allowable range of the meter.
And S7, when the meter reading is abnormal, carrying out manual verification, when the actual result of the meter is inconsistent with the result fed back by the meter reading recognition model, marking the image data, adding the sample image data, and repeating the steps S2 to S3 to train and update the meter reading recognition model.
When the number of a certain type of meters is large, the maximum measuring range and the minimum measuring range of the meters can be configured directly through the robot patrol configuration tool, and when the current reading of the meters is calculated, the maximum measuring range and the minimum measuring range in the configuration data are directly adopted, and the maximum measuring range and the minimum measuring range of the meters do not need to be identified by a model, so that the calculation amount can be greatly reduced, and meanwhile, a lot of human workload is not increased.
Claims (10)
1. A meter reading identification method based on a robot is characterized by comprising the following steps:
s1, collecting sample image data of the meter by the robot;
s2, cleaning and labeling the sample image data;
s3, training sample image data through a deep learning algorithm to obtain a meter reading recognition model;
s4, the robot patrols and collects the patrolling image data of the meter;
and S5, automatically calling a meter reading recognition model, and calculating to obtain the current reading of the meter.
2. The robot-based meter reading identification method of claim 1, wherein the step S1 includes the steps of:
s11, point location configuration is carried out on the meters needing automatic inspection in the factory area through the robot inspection configuration tool;
s12, the robot identifies and adjusts the pose of the robot and/or adjusts the direction of the camera according to the image, so that the center of the camera corresponds to the center of the meter;
and S13, the robot shoots at least one meter image and records corresponding robot position and posture data during shooting.
3. The robot-based meter reading identification method of claim 1, wherein the labeling in step S2 includes labeling a meter category.
4. The robot-based meter reading identification method of claim 3, wherein the labeling in step S2 includes labeling a meter maximum measurement range and a meter minimum measurement range.
5. The robot-based meter reading identification method of claim 1, wherein the cleaning of the sample image data in step S2 includes performing scaling, rotation, displacement, ray transformation, and image synthesis on the sample image data.
6. The robot-based meter reading identification method according to claim 1, wherein the most accurate image is selected as the patrol sample specimen image among the plurality of meter images when performing the deep learning in step S3.
7. The robot-based meter reading recognition method of claim 6, wherein the meter reading recognition model in the step S3 includes a meter type automatic recognition model for recognizing a meter type; when the robot patrols and examines, firstly, the meter type automatic identification model is called to identify the meter type, then the meter is photographed, and image data is collected.
8. The robot-based meter reading identification method of claim 7, wherein the meter reading identification model of step S5 detects straight lines by hough transform, extracts pointer images, fits the angles of the pointer and the maximum measurement range and the minimum measurement range by the least square method, and calculates the current meter reading by combining the identified maximum meter range and minimum meter range.
9. The robot-based meter reading identification method according to claim 8, wherein before the straight line is detected by using hough transform, the position of the robot holder is calibrated according to the image characteristics of the sample at the inspection point, so that the inspection image and the image acquisition position of the sample at the inspection point are consistent.
10. The robot-based meter reading identification method of any one of claims 1-9, further comprising:
s6, feeding the current meter reading back to the service management background, and checking the abnormal meter reading by the warning early warning module; the abnormal reading of the meter indicates that the current reading of the meter is not in the range allowed by the meter;
and S7, when the meter reading is abnormal, carrying out manual verification, when the actual result of the meter is inconsistent with the result fed back by the meter reading recognition model, marking the image data, adding the sample image data, and repeating the steps S2 to S3 to train and update the meter reading recognition model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111309307.6A CN114155432A (en) | 2021-11-06 | 2021-11-06 | Meter reading identification method based on robot |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111309307.6A CN114155432A (en) | 2021-11-06 | 2021-11-06 | Meter reading identification method based on robot |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114155432A true CN114155432A (en) | 2022-03-08 |
Family
ID=80458989
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111309307.6A Pending CN114155432A (en) | 2021-11-06 | 2021-11-06 | Meter reading identification method based on robot |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114155432A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114663680A (en) * | 2022-05-25 | 2022-06-24 | 天津大学四川创新研究院 | System and method for recognizing temperature and humidity meter readings |
-
2021
- 2021-11-06 CN CN202111309307.6A patent/CN114155432A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114663680A (en) * | 2022-05-25 | 2022-06-24 | 天津大学四川创新研究院 | System and method for recognizing temperature and humidity meter readings |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102521560B (en) | Instrument pointer image identification method of high-robustness rod | |
| CN103076589B (en) | A kind of digital multimeter automatic calibrator and calibration method thereof | |
| CN108764257A (en) | A kind of pointer instrument recognition methods of various visual angles | |
| CN113418543B (en) | Automatic driving sensor detection method and device, electronic equipment and storage medium | |
| CN105930837A (en) | Transformer station instrument equipment image recognition method based on autonomous routing inspection robot | |
| CN113688817A (en) | Instrument identification method and system for automatic inspection | |
| CN1804559A (en) | Water level meter with digital video frequency | |
| CN110953993A (en) | Detection device and method for sag and distance limit of power transmission line | |
| CN115294317A (en) | Pointer type instrument reading intelligent detection method for industrial production factory | |
| CN106705850A (en) | Image-based terminal quality detection method | |
| CN114155432A (en) | Meter reading identification method based on robot | |
| CN111289111A (en) | Self-calibration infrared body temperature rapid detection method and detection device | |
| CN112381190B (en) | Cable force testing method based on mobile phone image recognition | |
| CN111145247B (en) | Position degree detection method based on vision, robot and computer storage medium | |
| CN112489017A (en) | Intelligent identification method and system for power equipment faults | |
| CN116563391A (en) | Automatic laser structure calibration method based on machine vision | |
| CN116310263A (en) | Pointer type aviation horizon instrument indication automatic reading implementation method | |
| CN116245230A (en) | Operation inspection and trend analysis method and system for power station equipment | |
| CN113869112A (en) | Instrument automatic reading method and device based on machine vision | |
| CN203149105U (en) | Digital universal meter automatic calibrating device | |
| CN112858331A (en) | VR screen detection method and detection system | |
| CN110853080A (en) | Method for measuring size of field fruit | |
| CN109373899A (en) | A kind of complex parts geometric dimension precision measurement system and method | |
| CN112304430B (en) | Lighting site real-time intelligent monitoring method and device | |
| CN114719784B (en) | Device and method for detecting inclination angle of cylindrical weighing sensor |
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 |