A smart meter reading system and method

Abstract

The application discloses an intelligent meter reading system and method, wherein the system comprises node monitoring devices arranged at each power consumption node and a server arranged at a cloud end; the node monitoring device comprises an image acquisition module, a communication module, a power module and a main control module; the main control module is connected to the server through the communication module, is used for receiving data from the server and controlling the image acquisition module to acquire data; the image acquisition module is used for acquiring image data of a meter of the power consumption node and transmitting the data to the communication module; the server is used for receiving the image data from the node monitoring device, pre-processing the image data, outputting the image data as text data after text recognition, and converting the text data into power consumption data. The application realizes meter reading of each monitoring node independently by using image recognition technology and internet calculation.

Description

Technical Field

[0001] This invention belongs to the field of power system technology, specifically relating to an intelligent meter reading system and method. Background Technology

[0002] Remote meter reading systems are the result of the deep integration of modern computer, microelectronics, and network communication technologies. They can not only automatically measure, control, and store relevant physical quantities like conventional instruments, and display measurement results and control status, but also have network application characteristics: through the Internet, remote meter reading systems can be operated, measurement results can be obtained, parameters can be set, real-time monitoring and fault diagnosis can be performed, and dynamic information can be published on the Internet.

[0003] In existing technologies, electricity consumption in a region is often collected and summarized by a concentrator. However, the concentrator is usually installed in the region in the form of an external box, which is costly and requires complex maintenance. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide an intelligent meter reading system and method to reduce meter reading costs and improve meter reading accuracy.

[0005] To solve the above-mentioned technical problems, the present invention provides an intelligent meter reading system, comprising: a node monitoring device installed at each power consumption node and a server installed in the cloud, wherein the node monitoring device includes an image acquisition module, a communication module, a power supply module and a main control module;

[0006] The main control module is connected to the server via the communication module, and is used to receive data from the server and control the image acquisition module to acquire data.

[0007] The image acquisition module is used to acquire image data of the electricity meter at the electricity consumption node and transmit the data to the communication module;

[0008] The server receives image data from the node monitoring device, preprocesses and performs text recognition on the image data, outputs it as text data, and converts the text data into electricity consumption data.

[0009] Furthermore, the server has an image preprocessing module and a text recognition module. After receiving image data, the server preprocesses the image data through the image preprocessing module, which specifically includes binarization and noise reduction. The text recognition module is used to segment the preprocessed image data into individual characters, and then match and recognize the individual characters with numbers, characters, and letters in the database, outputting the recognized text data.

[0010] Furthermore, the intelligent meter reading system also includes a local storage module and a display module. The local storage module is used to store the image data collected each time, and the display module is used to call up and display the image data stored in the local storage module.

[0011] Furthermore, the smart meter reading system also includes a meter concentrator and a power metering unit. Multiple power metering units are connected to each power consumption node, the meter concentrator is connected to the meter of each power consumption node, and the power metering unit is connected to the local storage module.

[0012] This invention also provides a smart meter reading method, comprising:

[0013] Step S1: Acquire image data of a certain monitoring node;

[0014] Step S2: Preprocess the acquired image data, segment the characters in the image data, and recognize them using multiple text recognition models;

[0015] Step S3: If multiple text recognition models produce the same result, output the text data.

[0016] Step S4: If the recognition result of any one of the multiple text recognition models is different from the others, perform a first self-test on the text recognition model and output the text data after the first self-test.

[0017] Step S5: After adding data identifiers to the text data, output the power consumption data.

[0018] Further, step S2 specifically includes: preprocessing the image data, including binarization and noise reduction, and then extracting features from it; after feature extraction, segmenting the data characters in a fixed area character by character and then performing text recognition.

[0019] Furthermore, the specific steps of the first self-detection include:

[0020] Extract the first image data from the historical data collection where the recognition results of each text recognition model are the same and assign values ​​to it;

[0021] Assign values ​​to the second image data with different recognition results;

[0022] Calculate the average value of the first and second image data after assignment;

[0023] If the sample mean is within the preset confidence interval, text data is output; otherwise, a second self-test is performed until it passes and then text data is output.

[0024] Furthermore, the specific steps of the second self-detection include:

[0025] After acquiring the third image data of the current monitoring node and marking it as questionable data, it is transmitted to the server;

[0026] The system receives query response data from the server and simultaneously collects fourth image data; while collecting third and fourth image data, the metering unit simultaneously collects first and second electricity consumption data.

[0027] After removing the questioning model, multiple text recognition models are used to identify the third and fourth image data, and the recognition results are compared with the first and second power data.

[0028] If the recognition results of a text recognition model match, the text recognition model is marked as a confidence model and the text data is output; otherwise, it is marked as a questionable model.

[0029] Obtain the identification results of all questioning models, compare them with the first and second power data, and perform model correction.

[0030] Furthermore, the specific steps for correcting the questioning model include:

[0031] The text recognition model is compared sequentially with each questioning model based on the time when the model is questioned.

[0032] If the recognition result in this challenge model is a character shape error, and the total number and types of character shape errors are within the preset range, adjust the incorrectly identified character shape in the challenge model to the correct character shape before outputting it, and delete the challenge mark of the challenge model; otherwise, delete the text recognition model.

[0033] Furthermore, the first self-detection and the second self-detection are performed at predetermined intervals.

[0034] The present invention has the following beneficial effects: by using image recognition and Internet computing, data images are collected by a camera and preprocessed, and the preprocessed images are recognized as text using multiple models. The validity of the text recognition results is judged by comparing the consistency of the recognition results of multiple text models. When the recognition results are inconsistent, the present invention uses a first self-detection and a second self-detection mechanism to realize error correction in the meter reading process, thus ensuring the accuracy of image recognition using the Internet. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0036] Figure 1 This is a schematic diagram of the structure of an intelligent meter reading system according to an embodiment of the present invention.

[0037] Figure 2 This is a flowchart illustrating a smart meter reading method according to Embodiment 2 of the present invention.

[0038] Figure 3 This is a schematic diagram of the specific process of a smart meter reading method according to Embodiment 2 of the present invention. Detailed Implementation

[0039] The following description of the embodiments is taken with reference to the accompanying drawings, which illustrate specific embodiments in which the invention can be implemented.

[0040] Please refer to Figure 1 As shown, Embodiment 1 of the present invention provides an intelligent meter reading system, including: a node monitoring device installed at each power consumption node and a server installed in the cloud. The node monitoring device includes an image acquisition module, a communication module, a power supply module and a main control module.

[0041] The main control module is connected to the server via the communication module, and is used to receive data from the server and control the image acquisition module to acquire data.

[0042] The image acquisition module is used to acquire image data of the electricity meter at the electricity consumption node and transmit the data to the communication module;

[0043] The server receives image data from the node monitoring device, preprocesses and performs text recognition on the image data, outputs it as text data, and converts the text data into electricity consumption data.

[0044] As a preferred embodiment, the image acquisition module of this embodiment can be a camera with a photo-taking function. The camera captures the readings of a mechanical or digital meter, and the acquired image data is transmitted to the server through the communication module for processing.

[0045] The server has an image preprocessing module and a text recognition module. After receiving image data, the server first preprocesses the image data using the image preprocessing module. Image preprocessing includes binarization and noise reduction. After preprocessing, the image data highlights the data content, weakens the background, and removes noise. The text recognition module first performs character segmentation on the preprocessed image, that is, it decomposes the data in the recognition area one by one to obtain individual characters. Then, it matches and recognizes each individual character with numbers, characters, and letters in the database, and outputs the recognized text data. The communication module is responsible for receiving control commands from the server and transmitting monitoring data displayed in text form to the server. In this embodiment, the text recognition module adopts common text recognition technology in the prior art, which will not be described in detail here.

[0046] In a preferred embodiment, the smart meter reading system further includes a local storage module and a display module. The local storage module stores the image data collected each time, and the display module retrieves and displays the image data stored in the local storage module. Furthermore, the smart meter reading system includes a meter concentrator and power metering units. Multiple power metering units are connected to each electricity consumption node, the meter concentrator is connected to the meter at each electricity consumption node, and the power metering units are connected to the local storage module. The meter concentrator and power metering units are common devices for collecting electricity data in the prior art. In this embodiment, the meter concentrator and power metering units serve as error correction mechanisms; when the electricity collected by the data acquisition unit is incorrect, the power metering units correct the error, rather than being the primary means of meter reading.

[0047] Corresponding to the intelligent meter reading system described in Embodiment 1 of the present invention, Embodiment 2 of the present invention also provides an intelligent meter reading method, such as... Figure 2 As shown, it includes the following steps:

[0048] Step S1: Acquire image data of a certain monitoring node;

[0049] Step S2: Preprocess the acquired image data, segment the characters in the image data, and recognize them using multiple text recognition models;

[0050] Step S3: If multiple text recognition models produce the same result, output the text data.

[0051] Step S4: If the recognition result of any one of the multiple text recognition models is different from the others, perform a first self-test on the text recognition model and output the text data after the first self-test.

[0052] Step S5: After adding data identifiers to the text data, output the power consumption data.

[0053] Specifically, please combine with Figure 3 As shown, after acquiring image data of the electricity meter at the monitoring node in step S1, the image data is transmitted to the server. Step S2 preprocesses the image data, including binarization and denoising, and then performs feature extraction. Feature extraction extracts character attribute features from the image that are beneficial for text recognition, including the geometric shape information of the characters, to facilitate subsequent text recognition. After feature extraction, the data characters within a fixed area are segmented character by character before text recognition is performed.

[0054] Since a single model cannot always identify the correct result, to ensure the reliability of the identification result, step S4 needs to apply a first self-test. The specific steps for performing the first self-test are as follows:

[0055] Extract the first image data from the historical data collection where the recognition results of each text recognition model are the same and assign values ​​to it;

[0056] Assign values ​​to the second image data with different recognition results;

[0057] Calculate the average value of the first and second image data after assignment;

[0058] If the sample mean is within the preset confidence interval, text data is output; otherwise, a second self-test is performed until it passes and then text data is output.

[0059] In this embodiment, three text recognition models are used as examples.

[0060] First, both Model A and Model B are used for recognition. The recognition results a and b are compared. If a = b, the result is output.

[0061] If a≠b, then model C is used for re-identification. The identification result c is compared with a and b. When c=a or c=b, the result c is output.

[0062] When a, b, and c are all unequal, indicating an error in the recognition result, the image is marked as p. i p i The value is assigned to 0. If at least two of a, b, and c are equal, meaning the recognition result is correct, then the first self-detection is correct, and the image is marked as q. j .

[0063] The marked p i q j Together they form a set R, that is:

[0064] R = (p i ,q j ), i = 1, 2, ..., m, j = 1, 2, ..., n

[0065] Then, calculate the recognition accuracy I for the number of samples within set R:

[0066]

[0067] Where I represents the recognition accuracy; m+n represents the total number of samples, where m is the value of p. i The number of q, n is q j Quantity; p i q j For image tag values.

[0068] The confidence interval for the sample is set according to the actual meter reading accuracy requirements. The recognition accuracy confidence interval is set to [e, f]. When I ≥ e, the accuracy of the recognition result is considered to be within the normal range, the collection and recognition process has not encountered any faults, and the next round of collection is directly entered; when I < f, the second self-test is performed.

[0069] The specific steps for performing the second self-test are as follows:

[0070] After acquiring the third image data of the current monitoring node and marking it as questionable data, it is transmitted to the server;

[0071] The system receives query response data from the server and simultaneously collects fourth image data; while collecting third and fourth image data, the metering unit simultaneously collects first and second electricity consumption data.

[0072] After removing the questioning model, multiple text recognition models are used to identify the third and fourth image data, and the recognition results are compared with the first and second power data.

[0073] If the recognition results of a text recognition model match, the text recognition model is marked as a confidence model and the text data is output; otherwise, it is marked as a questionable model.

[0074] Obtain the identification results of all questioning models, compare them with the first and second power data, and perform model correction.

[0075] The specific steps for correcting the questioning model are as follows:

[0076] The text recognition model is compared sequentially based on the time when it is challenged. If the recognition result of the challenged model is a character shape error, the incorrect character shape is adjusted to the correct character shape and output when the total number and types of character shape errors are within the preset range; otherwise, the text recognition model is deleted.

[0077] The term "character error" refers to a misjudgment between two similar characters, such as the numbers 1 and 7, which are easily misjudged on an LCD digital display screen. When the total number and types of character errors are within an acceptable preset range, the text recognition model is considered to have a training error. After deleting the challenge marker of the challenged model and correcting the incorrect character to the correct character, the text recognition model can continue to be used. Otherwise, the text recognition model cannot continue to be used and should be deleted from the server.

[0078] As a preferred embodiment, the server periodically adds new text recognition models to avoid the server becoming unusable due to excessive deletion of models after long-term use. At the same time, the more text recognition models there are, the more accurate the text recognition results will be.

[0079] As a preferred embodiment, when performing the second self-test, if the proportion of correct recognition in historical data and the proportion of recognition that was not questioned in this instance are both higher than the proportion of questioned models in all text models, when comparing the questioned models, the model directly compares the text recognition data of the correct text recognition model in this instance, rather than comparing the data of the electricity metering unit.

[0080] Finally, the identification results and electricity consumption data are output and stored in a unified format as electricity consumption text. The electricity consumption data is displayed in text form, which includes the time of collection of electricity consumption data, meter number, monitoring node number, electricity metering unit number, concentrator number, image acquisition module number and electricity consumption data. The corresponding image data is also stored when storing the electricity consumption text.

[0081] As a preferred embodiment, this embodiment performs the first and second self-tests at a predetermined cycle even when no first or second self-test occurs, and performs the first and second self-tests periodically to prevent errors in meter reading results.

[0082] As can be seen from the above description, compared with the prior art, the beneficial effects of the present invention are as follows: by using image recognition and Internet computing, data images are collected by a camera and preprocessed, and the preprocessed images are recognized as text using multiple models. The validity is judged by comparing the consistency of the recognition results of multiple text models. When the recognition results are inconsistent, the present invention uses the first self-detection and the second self-detection mechanism to realize the error correction of the meter reading process, thus ensuring the accuracy of image recognition using the Internet.

[0083] The above description is merely a preferred embodiment of the present invention and should not be construed as limiting the scope of the invention. Therefore, any equivalent variations made in accordance with the claims of the present invention are still within the scope of the present invention.

Claims

1. A smart meter reading method, characterized in that, include: Step S1: Acquire image data of a certain monitoring node; Step S2: Preprocess the acquired image data, segment the characters in the image data, and recognize them using multiple text recognition models; Step S3: If multiple text recognition models produce the same result, output the text data. Step S4: If the recognition result of any one of the multiple text recognition models is different from the others, perform a first self-test on the text recognition model and output the text data after the first self-test. Step S5: After adding data identifiers to the text data, output it as electricity consumption data; The specific steps of the first self-test include: Extract the first image data from the historical data collection where the recognition results of each text recognition model are the same and assign values ​​to it; Assign values ​​to the second image data with different recognition results; Calculate the average value of the first and second image data after assignment; If the sample mean is within the preset confidence interval, text data is output; otherwise, a second self-test is performed until it passes and then text data is output.

2. The intelligent meter reading method according to claim 1, characterized in that, Step S2 specifically includes: preprocessing the image data, including binarization and denoising, and then extracting features; after feature extraction, segmenting the data characters in a fixed area character by character and then performing text recognition.

3. The intelligent meter reading method according to claim 1, characterized in that, The specific steps of the second self-test include: After acquiring the third image data of the current monitoring node and marking it as questionable data, it is transmitted to the server; The system receives query response data from the server and simultaneously collects fourth image data; while collecting third and fourth image data, the metering unit simultaneously collects first and second electricity consumption data. After removing the questioning model, multiple text recognition models are used to identify the third and fourth image data, and the recognition results are compared with the first and second power data. If the recognition results of a text recognition model match, the text recognition model is marked as a confidence model and the text data is output; otherwise, it is marked as a questionable model. Obtain the identification results of all questioning models, compare them with the first and second power data, and perform model correction.

4. The intelligent meter reading method according to claim 3, characterized in that, The specific steps for correcting the questioning model include: The text recognition model is compared sequentially with each questioning model based on the time when the model is questioned. If the recognition result in this challenge model is a character shape error, and the total number and types of character shape errors are within the preset range, adjust the incorrectly identified character shape in the challenge model to the correct character shape before outputting it, and delete the challenge mark of the challenge model; otherwise, delete the text recognition model.

5. The intelligent meter reading method according to claim 4, characterized in that, The first self-test and the second self-test are performed at a predetermined cycle.

Images