Instrument reading visual identification method and device based on neural network algorithm

Through the visual recognition method of instrument readings based on neural network algorithm, the identification of scale and pointer key points through image preprocessing and area prediction models, the problems of low efficiency and low accuracy of instrument reading are solved, and the accuracy of automatic meter reading is achieved and efficient.

CN119068472APending Publication Date: 2024-12-03STATE GRID CORPORATION OF CHINA +1
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202411080099.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-07
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

In the prior art, instrument reading efficiency is low and inaccurate, and rely on manual reading. Deep learning methods are greatly affected by training samples and image quality, resulting in inaccurate readings.

Method used

The visual recognition method of instrument readings based on neural network algorithm is adopted to accurately identify instrument readings by collecting image information, image preprocessing, building training sample sets and area prediction models, identifying scales and pointer key points, and accurately identifying instrument readings.

Benefits of technology

The partitioned reading method reduces reading errors, increases the number of partitions to ensure reading accuracy, realizes automatic meter reading, and improves the accuracy and efficiency of readings.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119068472A_ABST
    Figure CN119068472A_ABST
Patent Text Reader

Abstract

The invention discloses an instrument reading visual identification method and device based on a neural network algorithm. The method comprises the following steps: S1, collecting image information of an instrument to be identified; s2, carrying out image preprocessing on the image information; s3, constructing a training sample set, collecting standard image information of a standard instrument, partitioning the standard image information, and then performing region division through scale key points and pointer key points; s4, constructing a region prediction model, and training the region prediction model by using the divided training sample set; and S5, carrying out scale key point and pointer key point identification and scale and pointer region prediction on the preprocessed image information in the step S2 by using the trained region prediction model, then using instrument reading identification according to the output region, and outputting a reading value.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to a method and device for visually identifying instrument readings based on a neural network algorithm. Background Art

[0002] Automated meter reading is a widely used method in measurement system applications, such as water meter reading and billing. Meter reading is also used in monitoring systems.

[0003] At present, most meter readings still need to be read manually, which is time-consuming, labor-intensive and inefficient. Computer vision technology and deep learning technology have made great progress in recent years. The use of computer vision to realize automatic meter reading to replace manual meter reading is increasingly used in the field of smart meters. However, deep learning usually relies on the training of a large number of samples. The training sample images directly affect the accuracy of deep learning readings. At the same time, they are also easily affected by image quality, resulting in inaccurate readings and affecting normal use.

[0004] Therefore, in the face of the above-mentioned problems, it is a technical problem that needs to be solved urgently to develop an instrument reading visual recognition technology based on a neural network algorithm that can ensure the accuracy of the readings. Summary of the invention

[0005] The purpose of the present invention is to provide a method and device for visually recognizing instrument readings based on a neural network algorithm, so as to solve the problems of low efficiency and inaccurate readings of existing instruments and meters.

[0006] In order to solve the above technical problems, the present invention provides a method for visually recognizing instrument readings based on a neural network algorithm, comprising the steps of:

[0007] S1: Collect image information of instruments to be identified;

[0008] S2: perform image preprocessing on the image information;

[0009] S3: Construct a training sample set, collect standard image information of standard instruments and meters, partition the standard image information, and then divide the area by scale key points and pointer key points;

[0010] S4: construct a regional prediction model and train the regional prediction model using the divided training sample set;

[0011] S5: Use the trained area prediction model to identify the scale key points and pointer key points and predict the scale and pointer areas of the image information preprocessed in step S2, and then use the instrument readings to identify the output area and output the reading values.

[0012] Furthermore, a visual camera is used to collect image information of instruments and meters.

[0013] Furthermore, in step S1, the size of the collected image information of the instrument to be identified is unified as 480 pixels in height and 640 pixels in width.

[0014] Furthermore, in step S2, image preprocessing includes: performing image defogging and image enhancement processing on the image information.

[0015] Furthermore, the image enhancement processing includes: using an unsharp masking algorithm to perform image enhancement to enhance the details of useful information in the image.

[0016] Furthermore, in step S3, the standard image information is evenly partitioned at intervals of 5 to 25 degrees.

[0017] Furthermore, in step S3, the scale and pointer image features of the training sample set are obtained, and then the scale and pointer are divided into regions according to the angle between the straight line where the pointer is located and the center of the scale circle.

[0018] Furthermore, in step S5, the identification of scale key points includes: after detecting the scale, cropping the scale information and comparing it with the scale information in the sample set to confirm the scale information; the identification of pointer key points includes: after detecting the pointer, locating the area where the pointer center axis is located on the pointer area image cropped according to the detected position information.

[0019] In addition, the present invention also provides an instrument reading visual recognition device based on a neural network algorithm, comprising: a processor and a memory; wherein the memory is used to store computer execution instructions, and when the device is running, the processor executes the computer execution instructions stored in the memory to enable the device to execute the above-mentioned instrument reading visual recognition method based on the neural network algorithm.

[0020] The beneficial effects of the present invention are as follows: the reading error can be reduced by the partitioned reading method, the reading can be guaranteed to be within a certain range by increasing the number of partitions, and a secondary reading is performed on the partitioned image to ensure the accuracy of the reading, realize automatic meter reading, and be easy to promote and apply. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The same reference numerals are used in these drawings to represent the same or similar parts. The exemplary embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0022] Figure 1 The figure is a flow chart of an embodiment of the present invention. DETAILED DESCRIPTION

[0023] like Figure 1 A method for visually identifying instrument readings based on a neural network algorithm is shown, comprising the steps of:

[0024] S1: Collect image information of instruments to be identified;

[0025] S2: perform image preprocessing on the image information;

[0026] S3: Construct a training sample set, collect standard image information of standard instruments and meters, partition the standard image information, and then divide the area by scale key points and pointer key points;

[0027] S4: construct a regional prediction model and train the regional prediction model using the divided training sample set;

[0028] S5: Use the trained area prediction model to identify the scale key points and pointer key points and predict the scale and pointer areas of the image information preprocessed in step S2, and then use the instrument readings to identify the output area and output the reading values.

[0029] The present invention can reduce reading errors through a partitioned reading method, ensure that the reading is within a certain range by increasing the number of partitions, and perform a secondary reading on the partitioned image, thereby ensuring the accuracy of the reading, realizing automatic meter reading, and being easy to promote and apply.

[0030] According to one embodiment of the present application, a visual camera is used to collect image information of instruments.

[0031] According to an embodiment of the present application, in step S1, the size of the image information of the instrument to be identified is unified to 480 pixels in height and 640 pixels in width. This embodiment can ensure the automatic reading effect by unifying the size of the image information.

[0032] According to an embodiment of the present application, in step S2, image preprocessing includes: performing image defogging and image enhancement on the image information. This embodiment can facilitate the use of some photos that are difficult to read through image preprocessing.

[0033] According to an embodiment of the present application, the image enhancement process includes: using an unsharp masking algorithm to perform image enhancement to enhance the details of useful information in the image. This embodiment can enhance the details of useful information in the image through image enhancement process.

[0034] According to an embodiment of the present application, in step S3, the standard image information is evenly partitioned at intervals of 5 to 25 degrees, and specifically, the standard image information can be partitioned at intervals of 5 degrees or 15 degrees.

[0035] According to an embodiment of the present application, in step S3, the scale and pointer image features of the training sample set are obtained, and then the scale and pointer are divided into regions according to the angle between the straight line where the pointer is located and the center of the scale circle.

[0036] According to one embodiment of the present application, in step S5, the identification of scale key points includes: after detecting the scale, cropping the scale information and comparing it with the scale information in the sample set to confirm the scale information; the identification of pointer key points includes: after detecting the pointer, locating the area where the pointer center axis is located on the pointer area image obtained by cropping the detected position information.

[0037] In addition, the present invention also discloses a visual recognition device for instrument readings based on a neural network algorithm, comprising: a processor and a memory; wherein the memory is used to store computer execution instructions, and when the device is running, the processor executes the computer execution instructions stored in the memory to enable the device to execute the above-mentioned visual recognition method for instrument readings based on a neural network algorithm.

[0038] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.

Claims

1. A method for visual recognition of instrument readings based on a neural network algorithm, characterized in that: Includes steps: S1: Collect image information of instruments to be identified; S2: performing image preprocessing on the image information; S3: construct a training sample set, collect standard image information of standard instruments and meters, partition the standard image information, and then divide the area by scale key points and pointer key points; S4: construct a regional prediction model and train the regional prediction model using the divided training sample set; S5: Use the trained area prediction model to identify the scale key points and pointer key points and predict the scale and pointer areas of the image information preprocessed in step S2, and then use the instrument readings to identify the output area and output the reading values.

2. The method for visual recognition of instrument readings based on a neural network algorithm according to claim 1, characterized in that: Visual cameras are used to collect image information of instruments.

3. The method for visual recognition of instrument readings based on a neural network algorithm according to claim 2 is characterized in that: In step S1, the size of the collected image information of the instrument to be identified is unified as 480 pixels in height and 640 pixels in width.

4. The method for visual recognition of instrument readings based on a neural network algorithm according to claim 1, characterized in that: In the step S2, the image preprocessing includes: performing image defogging and image enhancement processing on the image information.

5. The method for visually recognizing instrument readings based on a neural network algorithm according to claim 4 is characterized in that: The image enhancement process includes: using an unsharp mask algorithm to perform image enhancement to enhance the details of useful information in the image.

6. The method for visual recognition of instrument readings based on a neural network algorithm according to claim 1, characterized in that: In the step S3, the standard image information is evenly partitioned at intervals of 5 to 25 degrees.

7. The method for visually recognizing instrument readings based on a neural network algorithm according to claim 6, characterized in that: In step S3, the scale and pointer image features of the training sample set are obtained, and then the scale and pointer are divided into regions according to the angle between the straight line where the pointer is located and the center of the scale circle.

8. The method for visual recognition of instrument readings based on a neural network algorithm according to claim 1, characterized in that: In step S5, the identification of the scale key points includes: after detecting the scale, cutting the scale information and comparing it with the scale information in the sample set to confirm the scale information; the identification of the pointer key points includes: after detecting the pointer, locating the area where the pointer center axis is located on the pointer area image cut out according to the detected position information.

9. A visual recognition device for instrument readings based on a neural network algorithm, characterized in that: It is characterized in that it includes: a processor and a memory; wherein the memory is used to store computer execution instructions, and when the device is running, the processor executes the computer execution instructions stored in the memory, so that the device executes the instrument reading visual recognition method based on the neural network algorithm described in any one of claims 1-8.

Citation Information

Cited By

  • Water meter dial multi-mode identification method and system based on AI vision

    CN120635653A

  • A Multimodal Recognition Method and System for Water Meter Based on AI Vision

    CN120635653B