CN109635806B - Ammeter value identification method based on residual error network - Google Patents
Ammeter value identification method based on residual error network Download PDFInfo
- Publication number
- CN109635806B CN109635806B CN201811515767.2A CN201811515767A CN109635806B CN 109635806 B CN109635806 B CN 109635806B CN 201811515767 A CN201811515767 A CN 201811515767A CN 109635806 B CN109635806 B CN 109635806B
- Authority
- CN
- China
- Prior art keywords
- image
- ammeter
- neural network
- preprocessing module
- network
- 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
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
-
- 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
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Multimedia (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
An ammeter numerical value identification method based on a residual error network comprises an image acquisition device, an image preprocessing module and a convolutional neural network; the method comprises the following steps of S1: the image acquisition device acquires an ammeter image and transmits the ammeter image to the image preprocessing module; s2: the image preprocessing module is used for carrying out gray processing on the acquired ammeter image; s3: the image preprocessing module carries out smooth filtering on the electric meter image subjected to the graying processing to eliminate noise in the electric meter image; s4: and the image preprocessing module performs image sharpening processing on the electric meter image after the smoothing filtering. The method not only solves the problems of low manual reading efficiency, easy error, insufficient precision of the existing automatic meter identification method and the like, but also conforms to the new trend of intelligent development, can be applied to monitoring of power grid instrument equipment, and has good application prospect in engineering application of a power system.
Description
Technical Field
The invention relates to the field of precision instruments, in particular to an ammeter numerical identification method based on a residual error network.
Background
The pointer instrument is used as a measuring instrument, is widely applied to various fields in social production and life, and plays a role in importance. At present, the reading of the pointer instrument is mainly finished by manual interpretation, however, the method is greatly influenced by human factors and has poor reliabilityAnd the efficiency is low. In recent years, with the continuous development and improvement of digital image processing technology, the identification of pointer instruments has also achieved tremendous progress. Robert Sablatning et al studied the identification methods of water meters and dial indicators based on the image processing technique of machine vision. The pointer of the instrument can rotate 0-360 degrees on the dial plate, and the scales are uniformly distributed. The identification process mainly comprises the following steps: the dial on the image is segmented by adopting an image segmentation algorithm, the deflection angle of the pointer is obtained by utilizing Hough transformation, and the reading is determined by an angle method. F.Correa Alegria et al [5-6] The pointer type meter automatic identification device has the advantages that the pointer type meter automatic identification device is researched by utilizing a machine vision technology, the applicable range of an automatic identification system is greatly expanded, and common power meters and the like can be effectively identified. They use silhouette method to obtain pointer image, and use Hough transformation to detect the included angle between two pointers to complete reading identification, and in addition, also propose automatic identification method for digital instrument. In summary, the positioning methods of the pointer and the scale mark at the present stage mainly include a silhouette method and a Hough transformation method, and most of the methods assume that the imaging environment of the instrument is ideal, that is, the dial is placed regularly, the optical axis of the camera is perpendicular to the plane of the dial, and illumination is uniform. The silhouette method has higher recognition accuracy when the illumination is sufficient, but has poor recognition accuracy when the environment is less ideal; when the Hough method processes thicker pointers, due to uneven illumination or interference of other factors, a certain deviation exists between a detected straight line and the center line of an actual pointer, and the recognition accuracy of an algorithm is affected. However, the environment of the actual test work is complex, for example: there are various types of dials, uneven illumination, reflection of glass on the surface of the meter, non-perpendicular optical axis of the camera with the plane of the dial, inclination of the dial, etc., which all cause difficulty in recognizing the image of the meter, and most recognition methods will generate larger errors or even fail.
Aiming at the defects of the existing meter identification method, the complex environment of the electric power inspection and maintenance ammeter is provided with an ammeter numerical identification method based on a residual error network, the powerful capability of a convolutional neural network in deep learning to image processing is utilized, the acquired dial image information is intelligently identified through designing a multi-layer convolutional neural network, and the digital information of the meter is identified. The intelligent meter value recognition system based on deep learning mainly comprises two steps of offline training and online prediction. For offline training, the convolutional neural network is trained through the dial images and the scales of the dial manually collected by histories. For online prediction we use a trained network for meter identification. During implementation, firstly, image information of the digital dial of the ammeter is collected through a camera on the wearable equipment, then the collected image information is preprocessed, and finally the image after prediction processing is identified through a multi-layer neural network, so that the purpose of obtaining the correct ammeter is achieved.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides an ammeter numerical value identification method based on a residual error network, which comprises the following specific technical scheme:
an ammeter numerical identification method based on a residual error network is characterized by comprising the following steps of: the system comprises an image acquisition device, an image preprocessing module and a convolutional neural network;
the steps of the method are adopted in the following steps,
s1: the image acquisition device acquires an ammeter image and transmits the ammeter image to the image preprocessing module;
s2: the image preprocessing module is used for carrying out gray processing on the acquired ammeter image;
s3: the image preprocessing module carries out smooth filtering on the electric meter image subjected to the graying processing to eliminate noise in the electric meter image;
s4: the image preprocessing module performs image sharpening processing on the electric meter image after smooth filtering;
s5: the image preprocessing module converts the sharpened image into a binary image;
s6: the method comprises the steps that a training data set and a label set are arranged, N ammeter images stored through manual meter reading are collected, scale values corresponding to dial images in each ammeter image are used as label data, and the scale values are added into the label set;
preprocessing N ammeter images in sequence according to the steps 2 to 5, and adding the preprocessed ammeter images into a data set, wherein tag data in the tag set corresponds to data in the data set;
s7: building a convolutional neural network, and training the neural network in the following training mode:
the data in the dataset is taken as input to a neural network,
setting f all And theta (theta) all The method is respectively set as an identification function and all parameters of the whole neural network, and the expression of the whole neural network is as follows:
y＝f all (IMG,Θ all )
the input of the neural network is IMG, the IMG is an original image of an input ammeter dial, the original image is a dial image stored by manual meter reading, the label data is a scale value corresponding to the original image of the ammeter dial, and y is the output of the whole neural network;
the mean square error expression of the whole network is obtained as
Training the neural network by minimizing the above, wherein label is a label vector which corresponds to the scale of the dial;
s8: and (3) inputting the preprocessed pictures in the data set to be processed into a convolutional neural network as input parameters, wherein the output of the convolutional neural network is the electric meter reading.
Further: the neural network is set to have a learning rate of 0.01, the iteration step number is 1000, the batch processing size is set to be 200, and the ADAM algorithm is used for continuously iteratively reducing the loss function value of the network to update the parameters of the network.
Further: the sharpening method in the step S4 adopts Laplacian sharpening.
The beneficial effects of the invention are as follows: the pointer instrument in the power grid inspection is taken as a research object, and the characteristic that the convolutional neural network has high-efficiency processing on the image is utilized in combination with the image processing technology and the deep learning knowledge, so that the automatic identification method of the pointer instrument is provided. The method automatically identifies the preprocessed image through the multilayer convolutional neural network, solves the problems of low manual reading efficiency, easy error, insufficient precision of the existing automatic meter identification method and the like, conforms to the new trend of intelligent development, can be applied to monitoring of power grid instrument equipment, and has good application prospect in engineering application of a power system.
Drawings
Fig. 1 is a flow chart of the operation of the present invention.
Detailed Description
The preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby making clear and defining the scope of the present invention.
An ammeter numerical value identification method based on a residual error network comprises an image acquisition device, an image preprocessing module and a convolutional neural network;
the steps of the method are adopted in the following steps,
s1: the image acquisition device acquires an ammeter image and transmits the ammeter image to the image preprocessing module;
s2: the image preprocessing module is used for carrying out gray processing on the acquired ammeter image;
s3: the image preprocessing module carries out smooth filtering on the electric meter image subjected to the graying processing to eliminate noise in the electric meter image;
s4: the image preprocessing module performs image sharpening processing on the electric meter image after smooth filtering;
s5: the image preprocessing module converts the sharpened image into a binary image;
s6: the method comprises the steps that a training data set and a label set are arranged, N ammeter images stored through manual meter reading are collected, scale values corresponding to dial images in each ammeter image are used as label data, and the scale values are added into the label set;
preprocessing N ammeter images in sequence according to the steps 2 to 5, and adding the preprocessed ammeter images into a data set, wherein tag data in the tag set corresponds to data in the data set;
s7: building a convolutional neural network, and training the neural network in the following training mode:
the data in the dataset is taken as input to a neural network,
setting f all And theta (theta) all The method is respectively set as an identification function and all parameters of the whole neural network, and the expression of the whole neural network is as follows:
y＝f all (IMG,Θ all )
the input of the neural network is IMG, the IMG is an original image of an input ammeter dial, the original image is a dial image stored by manual meter reading, the label data is a scale value corresponding to the original image of the ammeter dial, and y is the output of the whole neural network;
the mean square error expression of the whole network is obtained as
Training the neural network by minimizing the above, wherein label is a label vector which corresponds to the scale of the dial;
s8: and (3) inputting the preprocessed pictures in the data set to be processed into a convolutional neural network as input parameters, wherein the output of the convolutional neural network is the electric meter reading.
The specific working process is that the identified object is an ammeter, a wearable device with a deep learning automatic identification system is used for identifying the collected ammeter image, the wearable device automatically identifies the position pointed by the pointer of the dial plate and displays the degree of the pointed position, and finally the automatic ammeter identification function is completed. The system model is shown in fig. 1. The automatic identification system mainly comprises image acquisition and preprocessing and dial automatic identification based on deep learning.
The image acquisition and preprocessing process comprises the following steps:
the image acquisition is the first link for realizing the automatic interpretation of the wearable equipment, and the link mainly acquires the image of the pointer instrument in the actual production and living environment and transmits the image to the wearable equipment instrument for interpretation. There are many image capturing devices currently capable of completing the image capturing requirements in pointer instrument automatic identification systems, such as digital cameras, image capturing cards, digital image signal capturing cards, and the like. Meanwhile, in the image acquisition link, the better the acquired image quality is, the more accurate the later identification effect is. The good image should be clear, have higher resolution, the goal area of the image is clear with the border of the background area, and the background should be as simple as possible, interference factor is few, including the illumination is moderate too much. However, the requirements for the camera are correspondingly high and the price is very high in order to acquire images with high resolution. The image acquisition environments are various, some environments have complex background, some environments have strong or weak illumination, and the environments are frequently met in the practical application of the pointer instrument automatic identification system.
In the image preprocessing, the camera possibly contains noise in the acquired image due to various reasons such as illumination, climate and the like when the instrument image is acquired, and the accuracy of a final recognition result is affected. Therefore, the color image obtained from the camera needs to be preprocessed by an image processing method, useful information in the image is highlighted, interference information is removed, and the next dial positioning and pointer identification service is provided. The preprocessing of the pointer instrument image mainly comprises the following steps: graying of the image, smoothing of the image, enhancement of the image and binarization of the image.
And (5) graying the image. The image collected by the camera is a color image, contains a large amount of color information, and if the color image is directly processed, the operation is complex and the occupied storage space is large. Therefore, it is necessary to convert the color meter image into a gray scale image before performing meter pointer positioning. Although the gray-scale image contains only luminance information and no color information, it can reflect the distribution and characteristics of chromaticity and luminance level of the entire image, and can greatly reduce the amount of calculation of the subsequent processing of the meter image.
And (5) smoothing and filtering the image. Various noises are inevitably generated in the acquired meter image due to the influence of various factors in the image acquisition process. Noise greatly affects subsequent processing of the image, increases the occurrence probability of recognition errors, and causes recognition errors, so that filtering processing is required to be performed on the meter image to suppress noise and improve image quality. For a single image, the gray values of the noise points are obviously different from the gray values of surrounding pixels, and the influence of noise can be removed by utilizing the characteristic of the noise points through a filtering method.
And (5) sharpening the image. In the image preprocessing process, the image is usually subjected to smoothing treatment to achieve the purpose of eliminating noise, the edge of the image becomes blurred in the denoising process, the blurred image can influence the positioning and identifying effects of the instrument pointer, and the later edge detection is not facilitated, so that the instrument image is required to be sharpened, and the edge of the image can be sharpened. The most commonly used sharpening method is laplace sharpening, which can effectively improve image blurring due to diffusion effects.
And (5) binarizing the image. Image binarization is an important link of digital image processing, and has wide application in image segmentation.
The dial identification based on deep learning is designed as data collection, and a data set of ammeter identification needs to be prepared before the network is trained. Firstly, a large number of ammeter images shot by a camera are collected, some pictures with poor display effect are removed, the pictures are preprocessed by a 3 rd section image preprocessing method, then labels are added to samples so as to facilitate network learning, and the labels are set to be scale values pointed by dial pointers. We divide the preprocessed data set into training, validation and test sets. Here, our training set is 10000 ammeter pictures, the verification set is 1000 ammeter pictures, and the test set is 1000 ammeter pictures.
The dial identification based on the convolutional neural network is an algorithm of artificial intelligence, the characteristic extraction is mainly carried out on the identified image through the convolutional neural network in the learning process, the convolutional neural network is one of the most classical models in the deep learning, and the dial identification based on the convolutional neural network is an efficient identification algorithm widely applied to the fields of pattern identification, image processing and the like in recent years. The method skillfully utilizes less weight to achieve the effect that the fully connected network cannot achieve, and has the characteristics of simple structure, less training parameters, strong adaptability and the like, and the characteristics are more obvious when the input of the network is a multidimensional image. A convolutional network generally comprises a plurality of convolutional layers and a pooling layer, the convolutional neural network is subjected to multi-layer convolutional operation when an image is processed, and the learned characteristics are more comprehensive as the number of convolutional layers increases and the network 'understands' the image more deeply.
Firstly, a residual neutral network (ResNet) is adopted to identify the preprocessed image, the residual network comprises 3 convolution layers to extract the characteristics, a pooling layer is connected behind the residual layer to reduce the dimension of the characteristic vector output by the residual network, the pooled characteristic vector is input into a full-connection layer, and finally, the output of the full-connection layer is input into a neuron to obtain the final ammeter value.
The traditional convolution layer or full-connection layer has the problems of information loss, loss and the like more or less when information is transmitted, the network gradually deepens to challenge the counter-propagation capacity of the network, the number of layers is more and less, the gradient is smaller and smaller when the layers are propagated until the gradient disappears, and the training error is larger and larger along with the increase of the number of layers. ResNet solves these problems to some extent by allowing the original input information to be transmitted directly to later layers, protecting the integrity of the information, and the whole network only needs to learn the part of input and output differences, simplifying learning objectives and difficulty.
The convolution layer in ResNet is mainly used to initialize the input data and extract the picture feature values and is mainly composed of a series of parallel filters connected to the input signal by a set of weights, which compute the convolution along the frequency domain of the input. Typically, each filter in a CNN network processes data at a different time sequence, convolutionally summing the data through a sliding window. We describe herein a simple CNN network with a filter, assuming the convolution kernel size of the convolution filter as mxn, by sliding the filter over the data to be convolved, the MN data is weighted and summed to obtain the convolution output. The transformation formula for CNN is therefore:
where b is the bias, f (·) is the activation function, which as used herein is the relu activation function, expressed as: f (x) =max (0, x)
The first convolution layer in the residual error network comprises 2 convolution kernels, the size of each convolution kernel is 3 multiplied by 3, and 2 feature images are output by the input image through the first convolution layer; the second convolution layer comprises 4 convolution kernels, the size of each convolution kernel is 4 multiplied by 4, the image outputs 4 feature images through the second convolution layer, the third convolution layer comprises 1 convolution kernel, the size of each convolution kernel is 5 multiplied by 5, the image outputs 1 feature image through the third convolution layer, the output after the third convolution is added with the original input to obtain the output of a residual layer, the back of the residual layer is connected with a Maxpooling layer, the Maxpooling layer takes the maximum value in a window through a movable pooling window, and the change expression is shown as follows:
x * ＝max(x) (2)
the pooling layer is mainly used for reducing the calculation time and gradually establishing higher invariance of the space and the data structure, and representing the original picture characteristics through a smaller downsampling coefficient. The size of the pooling window is 3 multiplied by 3, and the size of the picture of the input image after the input image passes through the Maxpooling layer is changed into one third of the original size. Through the residual neural network and the pooling layer, we compress the picture and extract the representative feature vector, then input the fully connected layer to get the fully connected output, finally output the final dial by a neuron and using the relu activation functionScale value. The input layer is a preprocessed picture, which is herein the dial image of the electricity meter. To train the neural network, we learn the ownership weights and bias in the network using a point-to-point learning approach we will f all And theta (theta) all Respectively set as the identification function and all parameters of the whole neural network. The expression for the entire neural network is therefore:
y＝f all (IMG,Θ all ) (3)
the IMG is an originally input ammeter dial image, and y is the output of the whole neural network. The input of the neural network is dial images stored by manual meter reading, and the label data is the scale value corresponding to each dial image. All parameters in the network are updated by ADAM algorithm, so that we can obtain the Mean Square Error (MSE) expression of the whole network as
The label is a label vector, namely a real dial scale. We train the neural network by minimizing the above equation.
Thus, the dial identification system design based on deep learning is completed.
We validated the proposed method by experiment. We first collect historical dial data by hand
Then, inputting the preprocessed data set picture into a convolutional neural network for training, setting the learning rate to be 0.01, the iteration step number to be 1000 times, setting the batch processing bitch size to be 200, continuously iterating and reducing the loss function value of the network (namely the mean square error value of the network) to update the parameters of the network by using an ADAM algorithm, so that the whole training network is optimal, finally, using the trained convolutional neural network for identifying various types of electric meters, and a large number of test results show that the average identification accuracy can reach 90%.
Claims (2)
1. An ammeter numerical identification method based on a residual error network is characterized by comprising the following steps of: the system comprises an image acquisition device, an image preprocessing module and a convolutional neural network;
the steps of the method are adopted in the following steps,
s1: the image acquisition device acquires an ammeter image and transmits the ammeter image to the image preprocessing module;
s2: the image preprocessing module is used for carrying out gray processing on the acquired ammeter image;
s3: the image preprocessing module carries out smooth filtering on the electric meter image subjected to the graying processing to eliminate noise in the electric meter image;
s4: the image preprocessing module performs image sharpening processing on the electric meter image after smooth filtering;
s5: the image preprocessing module converts the sharpened image into a binary image;
s6: the method comprises the steps that a training data set and a label set are arranged, N ammeter images stored through manual meter reading are collected, scale values corresponding to dial images in each ammeter image are used as label data, and the scale values are added into the label set;
preprocessing N ammeter images in sequence according to the steps 2 to 5, and adding the preprocessed ammeter images into a data set, wherein tag data in the tag set corresponds to data in the data set;
s7: building a convolutional neural network, and training the neural network in the following training mode: the data in the dataset is taken as input to a neural network,
setting f all And theta (theta) all The method is respectively set as an identification function and all parameters of the whole neural network, and the expression of the whole neural network is as follows:
y＝f all (IMG,Θ all )
the input of the neural network is IMG, the IMG is an original image of an input ammeter dial, the original image is a dial image stored by manual meter reading, the label data is a scale value corresponding to the original image of the ammeter dial, and y is the output of the whole neural network;
the mean square error expression of the whole network is obtained as
Training the neural network by minimizing the above, wherein label is a label vector which corresponds to the scale of the dial;
s8: the method comprises the steps of inputting a preprocessed picture in a data set to be processed into a convolutional neural network as an input parameter, wherein the output of the convolutional neural network is an ammeter reading;
the convolutional neural network is set to have a learning rate of 0.01, the iteration step number is 1000, the batch processing size is set to be 200, and the ADAM algorithm is used for continuously iteratively reducing the loss function value of the network to update the parameters of the network.
2. The residual network-based ammeter value identification method as claimed in claim 1, wherein: the sharpening method in the step S4 adopts Laplacian sharpening.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811515767.2A CN109635806B (en) | 2018-12-12 | 2018-12-12 | Ammeter value identification method based on residual error network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811515767.2A CN109635806B (en) | 2018-12-12 | 2018-12-12 | Ammeter value identification method based on residual error network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109635806A CN109635806A (en) | 2019-04-16 |
| CN109635806B true CN109635806B (en) | 2023-08-25 |
Family
ID=66073129
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811515767.2A Active CN109635806B (en) | 2018-12-12 | 2018-12-12 | Ammeter value identification method based on residual error network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109635806B (en) |
Families Citing this family (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110334699A (en) * | 2019-05-31 | 2019-10-15 | 国网浙江省电力有限公司金华供电公司 | A kind of meter automatic recognition system based on deep learning |
| CN110688900A (en) * | 2019-08-27 | 2020-01-14 | 国网河北省电力有限公司电力科学研究院 | Withdrawal meter management method based on image recognition |
| CN111046943A (en) * | 2019-12-09 | 2020-04-21 | 国网智能科技股份有限公司 | Method and system for automatically identifying state of isolation switch of transformer substation |
| CN111209798B (en) * | 2019-12-23 | 2023-05-23 | 广州网文三维数字技术有限公司 | Special pressure equipment instrument identification method based on neural network |
| CN111950396B (en) * | 2020-07-27 | 2024-05-14 | 江苏大学 | Meter reading neural network identification method |
| CN113705487B (en) * | 2021-08-31 | 2023-08-08 | 西南交通大学 | Precision workpiece identification and technological parameter association system and identification method |
| CN113743405A (en) * | 2021-09-07 | 2021-12-03 | 南方电网数字电网研究院有限公司 | Intelligent meter reading method and device for electric energy meter |
| CN114140678A (en) * | 2021-11-23 | 2022-03-04 | 北京东方国信科技股份有限公司 | Scale recognition method and device, electronic equipment and storage medium |
| CN114419636A (en) * | 2022-01-10 | 2022-04-29 | 北京百度网讯科技有限公司 | Text recognition method, device, equipment and storage medium |
| CN115222817B (en) * | 2022-08-31 | 2022-11-29 | 成都千嘉科技股份有限公司 | Pointer positioning method of pointer type pressure gauge |
| CN117115802B (en) * | 2023-10-25 | 2024-03-26 | 山东潍微科技股份有限公司 | Character wheel type water meter digital identification and processing method based on deep learning |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101079094A (en) * | 2007-04-30 | 2007-11-28 | 中国科学院合肥物质科学研究院 | Meter reading identification device for long distance automatic meter reading system |
| CN102881141A (en) * | 2012-10-11 | 2013-01-16 | 西安北斗星数码信息股份有限公司 | Wireless ammeter reading recognition system |
| CN106228270B (en) * | 2016-07-27 | 2020-11-10 | 广东工业大学 | Energy consumption prediction method and system for big data driven extrusion equipment |
-
2018
- 2018-12-12 CN CN201811515767.2A patent/CN109635806B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN109635806A (en) | 2019-04-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109635806B (en) | Ammeter value identification method based on residual error network | |
| CN106875381B (en) | Mobile phone shell defect detection method based on deep learning | |
| CN107229930B (en) | Intelligent identification method for numerical value of pointer instrument | |
| CN106875373B (en) | Mobile phone screen MURA defect detection method based on convolutional neural network pruning algorithm | |
| CN109583324A (en) | A kind of pointer meters reading automatic identifying method based on the more box detectors of single-point | |
| CN110543878A (en) | pointer instrument reading identification method based on neural network | |
| CN112818988B (en) | Automatic identification reading method and system for pointer instrument | |
| CN104112269B (en) | A kind of solar battery laser groove parameter detection method and system based on machine vision | |
| CN111598098B (en) | Water gauge water line detection and effectiveness identification method based on full convolution neural network | |
| CN113808180B (en) | Heterologous image registration method, system and device | |
| CN111563896B (en) | Image processing method for detecting abnormality of overhead line system | |
| CN111950396A (en) | Instrument reading neural network identification method | |
| CN115830004A (en) | Surface defect detection method, device, computer equipment and storage medium | |
| CN115775236A (en) | Surface tiny defect visual detection method and system based on multi-scale feature fusion | |
| CN113705564B (en) | Pointer type instrument identification reading method | |
| Hou et al. | A pointer meter reading recognition method based on YOLOX and semantic segmentation technology | |
| CN117315670A (en) | Water meter reading area detection method based on computer vision | |
| CN112529003A (en) | Instrument panel digital identification method based on fast-RCNN | |
| CN112381751A (en) | Online intelligent detection system and method based on image processing algorithm | |
| CN109711253A (en) | Ammeter technique for partitioning based on convolutional neural networks and Recognition with Recurrent Neural Network | |
| Ni et al. | Multi-meter intelligent detection and recognition method under complex background | |
| CN115761606A (en) | Box electric energy meter identification method and device based on image processing | |
| Dai et al. | Intelligent ammeter reading recognition method based on deep learning | |
| CN115588196A (en) | Pointer type instrument reading method and device based on machine vision | |
| CN116363522A (en) | Coastal zone reclamation sea change remote sensing monitoring method based on deep learning |
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 |