Low-power-consumption intelligent meter reading system based on AI vision

Abstract

The invention discloses a low-power-consumption intelligent meter reading system based on AI vision, and relates to the technical field of intelligent meter reading, and the system comprises a collection end which is used for being detachably clamped outside a water meter dial plate, forming a meter reading environment with the water meter dial plate, carrying out the illumination intensity detection of the meter reading environment, and sending the detection result to a display end when the brightness of the current meter reading environment meets the requirement; and local analysis and identification are performed on the acquired instrument image through the embedded deep learning model, and an optimized numerical value is acquired and transmitted to the server. According to the invention, power consumption and cost caused by image storage and transmission are saved, off-line work is realized, dependence on network background and far-end computing power is avoided, immediate feedback and continuous monitoring are realized, abnormal conditions are found in time, energy management efficiency is improved, flow consumption and data storage requirements are greatly reduced, and the method is suitable for large-scale popularization and application. The system has the advantages of high accuracy, low power consumption, high flexibility, real-time monitoring, high cost effectiveness and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent meter reading, and in particular to a low-power intelligent meter reading system based on AI vision. Background Art

[0002] With the development of communication technology, high-speed networks have made cloud data interaction possible. Water meter data collection terminal devices can send data to cloud servers through wireless networks (such as 4G, 5G, Wi-Fi, etc.). These transmission technologies can meet the rapid upload of large amounts of data, allowing water meter images to reach the cloud processing center in a timely manner for processing and identification.

[0003] However, the current cloud-based meter reading has the following disadvantages:

[0004] 1) Data transmission depends on the network. If the network signal is not good, such as poor network coverage in remote areas or network congestion during peak hours, data transmission will be delayed. This may prevent the water meter data from being updated to the management system in a timely manner, affecting the timeliness of water fee calculation and water supply management. For example, in areas with weak signals such as basements in some old communities, data transmission by water meter data collection equipment may be frequently interrupted or delayed. At the same time, network instability may also cause data loss or transmission errors. Once a data packet is lost or erroneous during transmission, the water meter data received by the cloud will be inaccurate, affecting subsequent billing, statistics and other work.

[0005] 2) The operating cost is too high. To use cloud services, you need to pay fees to the cloud service provider. As the amount of data increases and the demand for computing resources grows, the cost will continue to rise. In the long run, this is a considerable expense, especially for large-scale urban water supply systems, where the storage and processing costs of numerous water meter data are high. At the same time, cloud recognition is highly dependent on cloud server resources. If there is a need for concurrent processing of large-scale water use data, the server computing resources may be tight. This will lead to a decrease in processing speed or even system crashes. For example, during the peak water use period in summer, a large amount of water meter data is uploaded and processed at the same time, which may overwhelm the cloud server. In addition, the operation of cloud servers consumes a lot of electricity, which is not conducive to sustainable development from an energy perspective. And as the scale of data centers expands, the problem of energy consumption becomes increasingly prominent.

[0006] Currently, no effective solution has been proposed for the problems in the related technologies. Summary of the invention

[0007] In view of the problems in the related technology, the present invention proposes a low-power intelligent meter reading system based on AI vision to overcome the above-mentioned technical problems existing in the existing related technology.

[0008] The technical solution of the present invention is achieved in this way:

[0009] A low-power intelligent meter reading system based on AI vision, comprising: an information interaction collection end and a service end, wherein;

[0010] The acquisition end is used to be detachably connected to the outside of the water meter dial and form a meter reading environment with the water meter dial, and detect the light intensity of the meter reading environment. When the brightness of the current meter reading environment meets the requirements, the acquired meter image is locally analyzed and recognized through the embedded deep learning model, and the optimized value is obtained and transmitted to the service end;

[0011] The server is used to store, analyze and process the received optimization values ​​and collection end tags, and generate reports and statistical data to provide decision support for administrators.

[0012] Furthermore, the acquisition end includes: a brightness acquisition module, a microcontroller module, an optical module, an image acquisition module, an AI visual processing module, a data transmission module and a power supply module, wherein;

[0013] The brightness acquisition module is used to be deployed in the acquisition terminal, and to collect the light intensity in the meter reading environment, and output the corresponding measurement value to the micro-control module;

[0014] The microcontroller module is used to obtain the measurement value of the brightness acquisition module and determine whether the measurement value of the current meter reading environment is greater than a preset brightness threshold value. If the measurement value of the current meter reading environment is greater than the preset brightness threshold value, it means that the brightness of the current meter reading environment meets the requirements, and a collection instruction is sent to the image acquisition module; if the measurement value of the current environment is less than the preset brightness threshold value, it means that the brightness of the current meter reading environment does not meet the requirements, and the optical module is controlled to perform fill light, and a collection instruction is sent to the image acquisition module after the fill light is performed;

[0015] The optical module is used to provide high-brightness fill light for the current meter reading environment;

[0016] The image acquisition module is used to start the high-definition camera according to a preset time interval or trigger instruction, capture the meter image of the water meter dial, and transmit the collected meter image to the AI ​​visual processing module;

[0017] The AI ​​visual processing module is used to analyze and identify the collected instrument images according to the pre-trained deep learning model, obtain the characteristic numbers on the instrument images, convert them into a readable data format, optimize the data format, and transmit the optimized values ​​to the data transmission module;

[0018] The data transmission module is used to bind the optimization value and the collection end tag and transmit them to the service end;

[0019] The power module is used to supply power to the acquisition end.

[0020] Furthermore, the collection end is provided with a plurality of light-transmitting holes arranged in an array, and the light-transmitting holes are connected to the meter reading environment for dew point supplementary lighting.

[0021] Furthermore, the acquisition end also includes: a sleep module, which is used to operate the acquisition end in a low power consumption mode during the current acquisition cycle and after the data transmission module transmits the optimized value to the server.

[0022] Furthermore, the brightness acquisition module includes: a light sensor.

[0023] Furthermore, the optical module includes: a line laser.

[0024] Furthermore, the step of obtaining characteristic numbers on the instrument image includes the following steps:

[0025] Pre-process the instrument image in advance to obtain the cropped image;

[0026] Extract the image containing the instrument digital features from the cropped image using the SIFT feature extraction algorithm;

[0027] The digital feature image of the instrument is input into the trained convolutional neural network. The convolutional neural network extracts features from the input digital feature image and compares and matches the extracted features with the features in the training data. Based on the matching results, the convolutional neural network outputs the recognition result, which includes the feature digits and converts the feature digits into a readable data format.

[0028] Further, the data format is optimized, including the following steps:

[0029] Convert the data format to string representation in advance;

[0030] Traverse the string from the left until the first non-zero character is found, and then intercept the part from this non-zero character to the end of the string, which is expressed as the optimized value.

[0031] Beneficial effects of the present invention:

[0032] 1. The collection end of the present invention is connected to the outside of the water meter dial by a detachable card, that is, the collection end module is directly installed on the original meter, without replacing the entire water meter, avoiding the high cost of replacing a new water meter. For a large number of water meter installation bases, only installing the collection end can save a lot of capital investment, especially for some old communities or water supply system renovation projects, there is no need to carry out large-scale water meter replacement projects, reducing construction costs and interference with residents' lives. At the same time, it reduces maintenance costs. The collection end is relatively independent. If a fault occurs, it can be repaired or replaced alone without processing the entire water meter, reducing the complexity and cost of maintenance.

[0033] 2. The present invention sets a plurality of light-transmitting holes arranged in an array at the collection end, and the light-transmitting holes are connected to the meter reading environment for dew point fill light to meet the daily assembly lighting needs. When the special environment provides high-brightness fill light for the current meter reading environment through the optical module, the details on the water meter dial can be made clearer, and the clarity of the meter image can be improved. At the same time, the high-brightness fill light can provide good meter image quality, so that the AI ​​visual processing module can more accurately identify the water meter readings, and the AI ​​visual processing module can operate under normal working parameters, reduce overload operation losses, extend service life, reduce energy consumption, and improve accuracy and reliability.

[0034] 3. The present invention starts a high-definition camera according to a preset time interval or trigger instruction, captures the meter image of the water meter dial, and transmits the collected meter image to the AI ​​visual processing module. At the same time, the AI ​​visual processing module analyzes and identifies the collected meter image according to a pre-trained deep learning model, obtains the characteristic numbers on the meter image, and converts them into a readable data format, and optimizes the data format, and transmits the optimized numerical transmission and the acquisition end label to the server end, so as to realize real-time reading of the water meter dial data, which not only saves the power consumption and cost caused by storing and transmitting images, but also realizes offline work, does not rely on the network background and remote computing power, realizes instant feedback and continuous monitoring, timely detects abnormal situations, improves the efficiency of energy management, and greatly reduces traffic consumption and data storage requirements. In addition, the low-power design can extend battery life, reduce energy consumption, meet the requirements of sustainable development, and has the advantages of high accuracy, low power consumption, strong flexibility, real-time monitoring and high cost-effectiveness. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0036] Figure 1 The figure is a schematic diagram of a low-power intelligent meter reading system based on AI vision according to an embodiment of the present invention.

[0037] In the figure:

[0038] 1. Collection end; 2. Server end;

[0039] 11. Brightness acquisition module; 12. Microcontroller module; 13. Optical module; 14. Image acquisition module; 15. AI visual processing module; 16. Data transmission module; 17. Sleep module; 18. Power module. DETAILED DESCRIPTION

[0040] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field belong to the scope of protection of the present invention.

[0041] According to an embodiment of the present invention, a low-power intelligent meter reading system based on AI vision is provided.

[0042] like Figure 1 As shown, the low-power intelligent meter reading system based on AI vision according to an embodiment of the present invention includes: a collection end 1 and a service end 2 for information interaction, wherein;

[0043] The collection terminal 1 is used to be detachably connected to the outside of the water meter dial and form a meter reading environment with the water meter dial. It also detects the light intensity of the meter reading environment. When the brightness of the current meter reading environment meets the requirements, it uses the embedded deep learning model to perform local analysis and recognition on the collected meter image, and obtains the optimized value to be transmitted to the server 2;

[0044] The server 2 is used to store, analyze and process the received optimization values ​​and labels of the collection terminal 1, and generate reports and statistical data to provide decision support for administrators.

[0045] In addition, the acquisition end 1 includes: a brightness acquisition module 11, a micro-control module 12, an optical module 13, an image acquisition module 14, an AI visual processing module 15, a data transmission module 16 and a power supply module 18, wherein;

[0046] The brightness acquisition module 11 is used to be deployed in the acquisition terminal 1, and to collect the light intensity in the meter reading environment, and output the corresponding measurement value to the micro-control module 12;

[0047] In this technical solution, the brightness acquisition module 11 can use a light sensor, and the light sensor is installed in the acquisition terminal 1, or integrated with the image acquisition module 14.

[0048] In addition, during application, the collection end 1 may be provided with a plurality of light-transmitting holes arranged in an array for dew point fill light.

[0049] The microcontroller module 12 is used to obtain the measurement value of the brightness acquisition module 11 and determine whether the measurement value of the current meter reading environment is greater than a preset brightness threshold.

[0050] If the measured value of the current meter reading environment is greater than the preset brightness threshold, it means that the brightness of the current meter reading environment meets the requirements, and a collection instruction is sent to the image collection module 14;

[0051] If the measured value of the current environment is less than the preset brightness threshold, it means that the brightness of the current meter reading environment does not meet the requirements, and the optical module 13 is controlled to perform fill light, and after the fill light is performed, a collection instruction is sent to the image collection module 14;

[0052] The technical solution specifically sets the brightness range of the meter reading environment of the water meter dial image to be within a reasonable range of 100 to 500 lux. However, in actual applications, the meter reading environment is blocked by the machine cover itself, so there is no situation where the brightness exceeds 500 lux. That is, the preset brightness threshold is set to 100 lux, which means that only the measurement value of the current meter reading environment is considered to be greater than 100 lux.

[0053] Optical module 13, used to provide high brightness fill light for the current meter reading environment;

[0054] In this technical solution, the optical module can adopt a line laser, use a single light source design, can work under harsh environmental conditions, have a wide-area divergence angle of 45° at a full angle, and the line laser illumination spot is uniform.

[0055] The image acquisition module 14 is used to start the high-definition camera according to a preset time interval or a trigger instruction, capture the meter image of the water meter dial, and transmit the collected meter image to the AI ​​visual processing module 15;

[0056] In this technical solution, for the above-mentioned high-definition camera, in order to improve the image quality, automatic focus can be adopted.

[0057] The AI ​​visual processing module 15 is used to analyze and identify the collected instrument image according to the pre-trained deep learning model, obtain the characteristic numbers on the instrument image, convert them into a readable data format, optimize the data format, and transmit the optimized values ​​to the data transmission module 16;

[0058] This technical solution, for the above-mentioned AI visual processing module 15, includes the following steps:

[0059] Pre-process the instrument image in advance to obtain the cropped image;

[0060] This technical solution includes the following steps:

[0061] Perform instrument image denoising, remove noise from instrument images through filtering algorithms, and improve the quality of instrument images;

[0062] Enhance the instrument image to increase its contrast and brightness, making the numbers on the instrument image clearer and more visible;

[0063] Perform instrument image cropping to generate a cropped image.

[0064] This technical solution adjusts the instrument image to a suitable size, which can reduce the amount of calculation and improve the recognition speed.

[0065] The cropped image is extracted using the SIFT (Scale-Invariant Feature Transform) feature extraction algorithm to obtain an image containing instrument digital features.

[0066] Specifically, the present technical solution, in order to extract the digital features of the instrument, converts the information in the cropped image into a set of feature vectors for subsequent recognition, extracts the digital feature image of the instrument, and can extract the shape, texture, and edge features of the digits.

[0067] The instrument digital feature image is input into the trained convolutional neural network (CNN). The convolutional neural network extracts features from the input digital feature image and compares and matches the extracted features with the features in the training data. Based on the matching results, the convolutional neural network outputs the recognition result, which includes the feature digits and converts the feature digits into a readable data format.

[0068] This technical solution, for optimizing the data format, includes the following steps:

[0069] Convert the data format to string representation in advance;

[0070] Traverse the string from the left until the first non-zero character is found, and then intercept the part from this non-zero character to the end of the string, which is expressed as the optimized value.

[0071] Data transmission module 16, used to bind the optimized value with the tag of collection terminal 1 and transmit it to service terminal 2;

[0072] In this technical solution, the data transmission module 16 can adopt wireless communication technology, such as NBIoT, 4GCat.1, and LoraWan, to reduce wiring costs and improve system flexibility.

[0073] In addition, the above-mentioned acquisition terminal 1 also includes:

[0074] The sleep module 17 is used to enable the collection terminal 1 to operate in a low power consumption mode during the current collection cycle and after the data transmission module 16 transmits the optimized value to the server 2 .

[0075] In the present technical solution, for the above-mentioned low power consumption mode, after the acquisition end 1 completes the acquisition, it enters the low power consumption mode to reduce power consumption and realize low power consumption operation. For example, the brightness acquisition module 11, the optical module 13, the image acquisition module 14 and the data transmission module 16 are turned off, and only the minimum power supply is reserved for the micro-control module 12 and the AI ​​visual processing module 15, so as to quickly wake up when needed.

[0076] The power module 18 is used to supply power to the acquisition terminal 1.

[0077] The server 2 is used to store, analyze and process the received optimization values ​​and labels of the collection terminal 1, and generate reports and statistical data to provide decision support for administrators.

[0078] In addition, as shown in Table 1, the low-power intelligent meter reading system based on AI vision of the present invention has the advantages of low operating power consumption, low data transmission cost and low cloud service cost compared with cloud recognition meter reading, as follows:

[0079]

[0080] Table 1 Performance comparison table

[0081] In summary, with the help of the above technical solution of the present invention, the following effects can be achieved:

[0082] 1. The collection terminal 1 of the present invention is connected to the outside of the water meter dial by a detachable card, that is, the collection terminal 1 module is directly installed on the original meter, without replacing the entire water meter, avoiding the high cost of replacing a new water meter. For a large number of water meter installation bases, only installing the collection terminal 1 can save a lot of capital investment, especially for some old communities or water supply system renovation projects, there is no need to carry out large-scale water meter replacement projects, reducing construction costs and interference with residents' lives. At the same time, it reduces maintenance costs. The collection terminal 1 is relatively independent. If a fault occurs, it can be repaired or replaced alone without processing the entire water meter, reducing the complexity and cost of maintenance.

[0083] 2. The present invention sets a plurality of light-transmitting holes arranged in an array through the collection end 1, and the light-transmitting holes are connected to the meter reading environment for dew point fill light to meet the lighting needs of daily assembly. When a special environment provides high-brightness fill light for the current meter reading environment through the optical module 13, the details on the water meter dial can be made clearer, and the clarity of the meter image can be improved. At the same time, the high-brightness fill light can provide good meter image quality, so that the AI ​​visual processing module 15 can more accurately identify the water meter reading, and the AI ​​visual processing module 15 can operate under normal working parameters, reduce overload operation losses, extend service life, reduce energy consumption, and improve accuracy and reliability.

[0084] 3. The present invention starts a high-definition camera according to a preset time interval or a trigger instruction, captures the meter image of the water meter dial, and transmits the collected meter image to the AI ​​visual processing module 15. At the same time, the AI ​​visual processing module 15 analyzes and identifies the collected meter image according to a pre-trained deep learning model, obtains the characteristic numbers on the meter image, and converts them into a readable data format, and optimizes the data format, and transmits the optimized numerical transmission and the acquisition end 1 label to the server 2, so as to realize real-time reading of the water meter dial data, which not only saves the power consumption and cost caused by storing and transmitting images, but also realizes offline work, does not rely on the network background and remote computing power, realizes instant feedback and continuous monitoring, timely detects abnormal situations, improves the efficiency of energy management, and greatly reduces traffic consumption and data storage requirements. In addition, the low-power design can extend battery life, reduce energy consumption, meet the requirements of sustainable development, and has the advantages of high accuracy, low power consumption, strong flexibility, real-time monitoring and high cost-effectiveness.

[0085] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. After considering the disclosure of the specification and the examples, those skilled in the art will easily think of other embodiments of the present disclosure. This application is intended to cover any modification, use or adaptation of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary technical means in the art that are not disclosed in the present disclosure. The specification and the examples are only regarded as exemplary, and the true scope and spirit of the present disclosure are indicated by the claims.

[0086] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A low-power intelligent meter reading system based on AI vision, characterized in that: include: The information exchange collection end (1) and the service end (2), wherein; The acquisition end (1) is used to be detachably connected to the outside of the water meter dial and form a meter reading environment with the water meter dial, and to detect the light intensity of the meter reading environment. When the brightness of the current meter reading environment meets the requirements, the acquired meter image is locally analyzed and identified through the embedded deep learning model, and the optimized value is obtained and transmitted to the service end (2); The server (2) is used to store, analyze and process the received optimization values ​​and labels of the collection end (1), and generate reports and statistical data to provide decision support for administrators.

2. The low-power intelligent meter reading system based on AI vision according to claim 1 is characterized in that: The acquisition end (1) comprises: a brightness acquisition module (11), a microcontroller module (12), an optical module (13), an image acquisition module (14), an AI visual processing module (15), a data transmission module (16) and a power supply module (18), wherein; The brightness acquisition module (11) is used to be deployed in the acquisition terminal (1), and to acquire the light intensity in the meter reading environment, and output the corresponding measurement value to the micro-control module (12); The microcontroller module (12) is used to obtain the measurement value of the brightness acquisition module (11) and determine whether the measurement value of the current meter reading environment is greater than a preset brightness threshold value, wherein if the measurement value of the current meter reading environment is greater than the preset brightness threshold value, it means that the brightness of the current meter reading environment meets the requirements, and a collection instruction is sent to the image acquisition module (14); if the measurement value of the current environment is less than the preset brightness threshold value, it means that the brightness of the current meter reading environment does not meet the requirements, and the optical module (13) is controlled to perform fill light, and after the fill light is performed, a collection instruction is sent to the image acquisition module (14); The optical module (13) is used to provide high-brightness supplementary light for the current meter reading environment; The image acquisition module (14) is used to start the high-definition camera according to a preset time interval or a trigger instruction, capture the meter image of the water meter dial, and transmit the captured meter image to the AI ​​visual processing module (15); The AI ​​visual processing module (15) is used to analyze and identify the collected instrument image according to the pre-trained deep learning model, obtain the characteristic numbers on the instrument image, convert them into a readable data format, optimize the data format, and transmit the optimized values ​​to the data transmission module (16); The data transmission module (16) is used to bind the optimized value and the tag of the collection end (1) and transmit it to the service end (2); The power supply module (18) is used to supply power to the acquisition end (1).

3. The low-power intelligent meter reading system based on AI vision according to claim 2 is characterized in that: The collection end (1) is provided with a plurality of light-transmitting holes arranged in an array, and the light-transmitting holes are connected to the meter reading environment for dew point supplementary light.

4. The low-power intelligent meter reading system based on AI vision according to claim 2 is characterized in that: The acquisition end (1) further comprises: a sleep module (17) for operating the acquisition end (1) in a low power consumption mode during a current acquisition cycle and after the data transmission module (16) transmits the optimized value to the server end (2).

5. The low-power intelligent meter reading system based on AI vision according to claim 2 is characterized in that: The brightness acquisition module (11) comprises: a light sensor.

6. The low-power intelligent meter reading system based on AI vision according to claim 2 is characterized in that: The optical module (13) comprises: a line laser.

7. The low-power intelligent meter reading system based on AI vision according to claim 1 is characterized in that: The method of obtaining the characteristic numbers on the instrument image comprises the following steps: Pre-process the instrument image in advance to obtain the cropped image; Extract the image containing the instrument digital features from the cropped image using the SIFT feature extraction algorithm; The digital feature image of the instrument is input into the trained convolutional neural network. The convolutional neural network extracts features from the input digital feature image and compares and matches the extracted features with the features in the training data. Based on the matching results, the convolutional neural network outputs the recognition result, which includes the feature digits and converts the feature digits into a readable data format.

8. The low-power intelligent meter reading system based on AI vision according to claim 7 is characterized in that: Optimizing the data format includes the following steps: Convert the data format to string representation in advance; Traverse the string from the left until the first non-zero character is found, and then intercept the part from this non-zero character to the end of the string, which is expressed as the optimized value.

Images