Water meter reading device and water meter reading method

Abstract

This invention provides a water meter reading device and a water meter reading method. The water meter reading device is suitable for installation on a water meter and includes a receiver and a processor. The receiver is used to collect audio signals, and the processor is coupled to the receiver. The processor converts the audio signals into a spectrum and detects multiple target feature blocks in the spectrum by inputting the spectrum into a neural network model. The processor counts the number of target feature blocks to determine the water usage of the water meter, wherein the number of target feature blocks reflects the number of rotations of the water meter's impeller.

Description

Technical Field

[0001] This invention relates to an automatic reading technology for instrument equipment, and more particularly to a water meter reading device and a water meter reading method. Background Technology

[0002] Water users install water meters to measure their water consumption. Traditionally, water meter readings are recorded by meter readers visiting the meter location in person. However, manual meter reading is not only labor-intensive but also inefficient. For water users living in remote areas, the cost and time involved in manual meter reading are even more significant. Furthermore, the water consumption readings recorded by meter readers can be incorrect due to human error.

[0003] With the development and advancement of communication technology, smart water meters have been proposed to alleviate the aforementioned difficulties. Through smart water meters (SWMs) and automatic meter reading systems (AMRs), water usage data from tap water users can be transmitted back in real time via the network, eliminating the need for manual meter reading. Specifically, current smart water meters have network capabilities and can transmit water usage data to a backend platform via wired or wireless transmission technologies to enable various applications. Currently, most automatic meter reading methods rely on computer vision technology. This method requires installing image acquisition devices on each water meter to take pictures, and then using image recognition technology to identify the water usage. However, image acquisition devices are prone to malfunction due to outdoor environmental factors such as sun and rain. Furthermore, when internal parts of the water meter malfunction and cannot accurately measure or display water usage, image recognition-based automatic meter reading methods cannot verify the accuracy of the water usage in real time, and may simply report incorrect readings. Summary of the Invention

[0004] In view of this, the present invention proposes a water meter reading device and a water meter reading method, which can obtain the water consumption of the water meter based on the audio signal generated by the rotation of the impeller inside the water meter.

[0005] This invention provides a water meter reading device, suitable for installation on a water meter, and includes a receiver and a processor. The receiver collects audio signals, and the processor is coupled to the receiver. The processor converts the audio signals into a spectrum and detects multiple target feature blocks in the spectrum by inputting the spectrum into a neural network model. The processor counts the number of target feature blocks to determine the water usage of the water meter, wherein the number of target feature blocks reflects the number of rotations of the water meter's impeller.

[0006] This invention provides a water meter reading method, applicable to water meter reading devices installed on water meters, comprising the following steps: Collecting audio signals using a radio receiver. Converting the audio signals into a spectrum. Detecting multiple target feature blocks in the spectrum by inputting the spectrum into a neural network model. Determining the water usage by counting the number of target feature blocks, wherein the number of target feature blocks reflects the number of rotations of the water meter's impeller.

[0007] Based on the above, in an embodiment of the present invention, the sound receiving device can receive audio signals at a location adjacent to the impeller of the water meter and convert the collected audio signals into a spectrum. By detecting target feature blocks in the spectrum of the audio signal using a trained neural network model, the target feature blocks reflected in one rotation of the impeller can be detected. Therefore, the number of rotations of the impeller inside the water meter can be determined by counting the target feature blocks in the spectrum, and the water usage rate of the water meter can be further determined accordingly. Thus, in an embodiment of the present invention, the water usage rate of the water meter can be determined based on the audio signal generated by the rotation of the impeller inside the water meter. Attached Figure Description

[0008] Figure 1 This is a block diagram of a water meter reading device according to an embodiment of the present invention;

[0009] Figure 2 This is a schematic diagram of a water meter reading device and a water meter according to an embodiment of the present invention;

[0010] Figure 3 This is a flowchart of a water meter reading method according to an embodiment of the present invention;

[0011] Figure 4 This is a schematic diagram illustrating the detection of the number of impeller rotations using a spectrum diagram according to an embodiment of the present invention;

[0012] Figure 5 This is a block diagram of a water meter reading device according to an embodiment of the present invention;

[0013] Figure 6 This is a flowchart of a water meter reading method according to an embodiment of the present invention;

[0014] Figure 7 This is a schematic diagram of a scenario in which multiple water meters are set up in the same field according to an embodiment of the present invention.

[0015] Explanation of reference numerals in the attached figures

[0016] 10,50,R1~RM: Water meter reading device;

[0017] N1: Network;

[0018] 20: Server equipment;

[0019] 110, 510: Radio receiver;

[0020] 120,520: Processor;

[0021] 130,530: Storage device;

[0022] 140,540: Transmission device;

[0023] 550: Image acquisition device;

[0024] 30: Water meter;

[0025] 31: Impeller;

[0026] 32: Degree display area;

[0027] S1: Spectrum diagram;

[0028] N1~N13: Target feature blocks;

[0029] Msg1: Warning message;

[0030] H1: Network repeater;

[0031] S301~S304, S601~S607: Steps. Detailed Implementation

[0032] Reference will now be made in detail to exemplary embodiments of the invention, examples of which are illustrated in the accompanying drawings. Wherever possible, the same component reference numerals are used in the drawings and description to denote the same or similar parts.

[0033] Figure 1 This is a block diagram of a water meter reading device according to an embodiment of the present invention. Please refer to... Figure 1 The water meter reading device 10 includes a receiver 110, a processor 120, a storage device 130, and a transmission device 140. The water meter reading device 10 is suitable for installation on a water meter with impeller-type metering. The water meter reading device 10 can be built into the water meter or attached to the outside of the water meter; this invention is not limited in this respect. Specifically, in some embodiments, the water meter reading device 10 and the water meter can be integrated into a single all-in-one smart water meter. Alternatively, in some embodiments, the water meter reading device 10 can be installed on a conventional water meter.

[0034] The recording device 110 is used to collect audio signals and includes a microphone or other electronic components capable of receiving audio signals. In some embodiments, the recording device 110 may also include an analog-to-digital converter, a filter, and an audio processor, etc.

[0035] Processor 120 is coupled to receiver 110. Processor 120 may be a central processing unit (CPU), or other programmable general-purpose or special-purpose microprocessor, digital signal processor (DSP), programmable controller, application-specific integrated circuit (ASIC), programmable logic device (PLD), or other similar device or combination of these devices.

[0036] Storage device 130 is coupled to processor 120 and is used to store audio data, instructions, program code, software components, and other data. It can be, for example, any type of fixed or removable random access memory (RAM), read-only memory (ROM), flash memory, hard disk, or other similar devices, integrated circuits, and combinations thereof. In some embodiments, processor 120 can access or execute program code, firmware / software modules, instructions, and other data recorded in storage device 130 to implement the water meter reading method of this invention.

[0037] The transmission device 140 is coupled to the processor 120 for transmitting data to an external device. In some embodiments, the water meter reader 10 can transmit water usage data to the server device 20 via the network N1 through the transmission device 140. The transmission device 140 may include components supporting one or more wired / wireless communication standards. For example, the transmission device 140 may be an electronic component supporting RS485, Long Term Evolution (LTE), 5G, Wi-Fi, LoRa, Bluetooth, or other communication standards; this invention is not limited thereto. That is, the transmission device 140 may include transceiver circuitry, an antenna, or a wired signal transmission port, etc.

[0038] Figure 2 This is a schematic diagram of a water meter reading device and a water meter according to an embodiment of the present invention. Please refer to... Figure 2In some embodiments, the water meter reading device 10 can be directly installed on a conventional water meter 30. Therefore, consumers do not need to replace the original water meter 30 or undertake extensive construction to enable the conventional water meter 30 to function as a smart water meter. It should be noted that the water meter 30 uses an impeller-type metering method to measure water consumption, and the water meter 30 includes an impeller 31. When water flows through the water meter 30, the water flow drives the impeller 31 to rotate. Thus, the water meter 30 can accumulate the number of rotations of the impeller 31 to determine the amount of water used.

[0039] It is worth mentioning that the impeller 31 generates a specific audio signal when it rotates. Therefore, based on the audio signal collected by the receiver 110, the water meter reading device 10 can analyze the audio signal collected by the receiver 110 to determine the rotation status of the impeller 31 and obtain the water usage reading of the water meter 30. In addition, in some embodiments, the water meter 30 may also include a calculation mechanism for counting the number of rotations of the impeller 31, and display the water meter reading determined by the calculation mechanism in the reading display area 32. Thus, the water meter reading device 10 can verify the water usage reading determined based on the audio signal with the reading displayed in the reading display area 32, thereby determining the accuracy and reliability of the two readings.

[0040] Figure 3 This is a flowchart of a water meter reading method according to an embodiment of the present invention. Please refer to... Figure 3 The method of this embodiment is applicable to the water meter reading device 10 of the above embodiment. The following describes the detailed steps of this embodiment in conjunction with the various components in the water meter reading device 10.

[0041] In step S301, the processor 120 collects audio signals using the receiver 110. Next, in step S302, the processor 120 converts the audio signals into a spectrogram. Specifically, the receiver 110 collects audio signals near the impeller of the water meter; therefore, the audio signals collected by the receiver 110 may include ambient sound components and sound components generated by the impeller rotation. The processor 120 can perform a series of operations on the analog time-domain audio signal to generate a corresponding spectrogram. The horizontal axis of the spectrogram represents time, the vertical axis represents frequency, and the color intensity of each feature point on the spectrogram represents the energy intensity of a specific frequency band. In other words, the spectrogram includes the time information, frequency information, and intensity information of the audio signal. Many useful pieces of information in the audio signal can be revealed through the spectrogram, such as volume level or frequency distribution range.

[0042] In some embodiments, after sampling the audio signal, the processor 120 can use Fast Fourier Transform (FFT) processing to convert the audio signal into a corresponding spectrogram. In some embodiments, to avoid ambient noise affecting the water meter reading results, the processor 120 may also perform noise reduction processing on the audio signal before converting it into a spectrogram. For example, the processor 120 may use a filter to perform noise reduction processing on the audio signal to filter out some ambient noise components. In addition, in order to accurately receive the sound generated by the impeller rotation and reduce the reception of ambient noise, the receiving device 110 may include a directional microphone.

[0043] In step S303, the processor 120 detects multiple target feature blocks in the spectrogram by inputting it into the neural network model. In step S304, the processor 120 counts the number of target feature blocks to determine the water usage of the water meter. Specifically, the spectrogram of the audio signal may include M*N feature points (M and N are positive integers), and the spectrogram can be considered as image data input into the neural network model. The processor 120 can use a trained neural network model to detect target feature blocks in the spectrogram. The neural network model may include a Convolutional Neural Network (CNN) model. For example, the neural network model may be R-CNN, Fast R-CNN, Faster R-CNN, YOLO, or SSD used for object detection in a CNN model, but the present invention is not limited thereto. Here, the number of target feature blocks may reflect the number of rotations of the water meter's impeller. In some embodiments, one target feature block represents one rotation of the impeller. In other words, each time the impeller of the water meter rotates once, a corresponding target feature block should appear on the spectrum based on the specific audio frequency generated by the impeller's rotation.

[0044] Therefore, in some embodiments, the receiver 110 can acquire audio signals within a preset time interval, and the processor 120 inputs the spectrum corresponding to the preset time interval into a neural network model to determine the number of impeller rotations within the preset time interval. The present invention does not limit the length of the preset time interval. For example, the processor 120 can count the number of impeller rotations every two minutes, that is, the time period of the spectrum used to detect the number of impeller rotations is two minutes. In this way, by continuously counting the number of target feature blocks and accumulating the number of impeller rotations, the processor 120 can determine the water usage of the water meter.

[0045] On the other hand, the trained neural network model is pre-constructed based on deep learning using a training dataset and can be stored in storage device 130. In other words, the model parameters of the trained neural network model (such as the number of network layers and the weight information of each network layer) have been determined by pre-training and stored in storage device 130. Specifically, when the spectrogram is input to the neural network model, feature vectors (or feature maps) can be generated by convolution processing to extract features. These feature vectors are then fed into the classifier in the neural network model to determine whether the target feature block has been detected.

[0046] It should be noted that, in order to construct a neural network model that can detect target feature blocks from spectrograms, the training dataset used to train the neural network model includes many sample spectrograms. The generation of these sample spectrograms can be illustrated as follows: A recording device is used to record audio from one or more test water meters in a metering operation, and a series of audio processing steps are performed based on these recordings to generate many sample spectrograms. It is important to note that these sample spectrograms are generated under conditions of different water flow velocities, that is, under conditions of different impeller rotation speeds.

[0047] Furthermore, during the model training phase, the target feature blocks corresponding to different impeller speeds in these sample spectrograms have been selected and labeled. It is known that the block size and internal feature information of the target feature blocks corresponding to different impeller speeds will differ. Finally, these sample spectrograms are input one by one into the neural network model, and the error is calculated by comparing the detection results generated by the neural network model based on the sample spectrograms with the true information. This error is then used to train the neural network model using backpropagation. The error calculation method (i.e., the loss function) can be, for example, the squared difference or Softmax. In this way, the trained neural network model can be used to detect the target feature blocks corresponding to different impeller speeds.

[0048] Figure 4 This is a schematic diagram illustrating the detection of impeller rotation count using a spectrogram according to an embodiment of the present invention. Please refer to... Figure 4 Assume that processor 120 can acquire a spectrum S1 corresponding to a preset time interval Δt. Processor 120 can use a trained neural network model to detect multiple target feature blocks N1 to N13 from spectrum S1. Then, processor 120 can determine that the impeller has rotated 13 times within the preset time interval Δt based on the number of target feature blocks N1 to N13. In this way, by continuously using the trained neural network model to detect subsequent spectrums and accumulate the number of impeller rotations, processor 120 can accumulate water usage based on the cumulative number of impeller rotations. For example, whenever the number of impeller rotations accumulates to P times, the water usage increases by 1 degree.

[0049] Figure 5 This is a block diagram of a water meter reading device according to an embodiment of the present invention. Please refer to... Figure 5 The water meter reading device 50 includes a microphone 510, a processor 520, a storage device 530, a transmission device 540, and an image acquisition device 550. In this embodiment, the microphone 510, processor 520, storage device 530, and transmission device 540 are similar to... Figure 1 The radio receiver 110, processor 120, storage device 130, and transmission device 140 of the embodiment will not be described in detail here. The difference is that the water meter reading device 50 of this embodiment also includes an image acquisition device 550.

[0050] Image acquisition device 550 is used to capture images and includes a camera lens with a lens and a photosensitive component. The photosensitive component is used to sense the intensity of light entering the lens, thereby generating an image. The photosensitive component can be, for example, a charge-coupled device (CCD), a complementary metal-oxide-semiconductor (CMOS) device, or other components, and the present invention is not limited thereto. Image acquisition device 550 is used to capture the readings of the water meter's display area (e.g., ...). Figure 2 The degree display area 32 shown is used to take a picture to obtain a degree image.

[0051] Figure 6 This is a flowchart of a water meter reading method according to an embodiment of the present invention. Please refer to... Figure 6 The method of this embodiment is applicable to the water meter reading device 50 of the above embodiment. The following describes the detailed steps of this embodiment in conjunction with the various components in the water meter reading device 50.

[0052] In step S601, the processor 520 collects audio signals using the microphone 510. Next, in step S602, the processor 520 converts the audio signals into a spectrogram. In step S603, the processor 520 detects multiple target feature blocks in the spectrogram by inputting it into a neural network model. In step S604, the processor 520 counts the number of target feature blocks to determine the water usage amount of the water meter. Steps S601 to S604 are similar to steps S301 to S304, and will not be described again here.

[0053] In step S605, the processor 520 uses the image acquisition device 550 to capture an image of the water meter's reading display area. In step S606, the processor 520 obtains the displayed reading of the water meter based on the reading image and performs a verification procedure based on the displayed reading and the water usage reading. The processor 520 can identify the displayed reading of the water meter using optical character recognition (OCR) technology. It is understood that if the water meter is operating normally and the water usage reading determined based on the audio signal is not significantly deviated due to external environmental factors, the difference between the displayed reading based on image recognition and the water usage reading determined based on the audio signal should be less than a threshold value. Conversely, if the difference between the displayed reading based on image recognition and the water usage reading determined based on the audio signal is too large, it indicates that the water meter is not operating normally or that the water usage reading determined based on the audio signal is significantly deviated due to external environmental factors. Therefore, in this embodiment, the processor 520 can detect whether the above-mentioned abnormal phenomenon has occurred by performing a verification procedure based on the displayed reading and the water usage reading.

[0054] In one embodiment, if the difference between the displayed temperature and the water usage value is greater than a threshold, the processor 520 determines that the displayed temperature and water usage value have failed verification. If the difference between the displayed temperature and the water usage value is not greater than the threshold, the processor 520 determines that the displayed temperature and the water usage value have passed verification. Specifically, when the difference between the displayed temperature and the water usage value is greater than the threshold and the displayed temperature is greater than the water usage value determined based on the audio signal, it indicates that the collected audio signal may be affected by environmental noise or other external factors, causing the processor 520 to be unable to accurately estimate the number of impeller rotations based on the audio signal. On the other hand, when the difference between the displayed temperature and the water usage value is greater than the threshold and the water usage value determined based on the audio signal is greater than the displayed temperature, it indicates that the water meter's operation may be abnormal, causing the impeller's rotation to fail to properly drive the water meter's display area to show the corresponding water meter value.

[0055] In step S607, if the displayed water meter reading and the water usage reading fail to pass verification, the processor 520 issues a warning message via the transmission device 540. The processor 520 can report the warning message to the water user's handheld device, another water meter reader located in the same area, or a backend server. Thus, through mutual verification of the two automatic reading methods, this embodiment of the invention can improve the accuracy and reliability of automatic water meter reading.

[0056] Figure 7 This is a schematic diagram illustrating a scenario where multiple water meters are installed in the same area according to an embodiment of the present invention. Please refer to... Figure 7Multiple water meter readers R1 to RM can be centrally located on the roof of the building. These water meter readers R1 to RM can be linked to a nearby network repeater H1, which can then integrate the data provided by these water meter readers R1 to RM and transmit it to the server device 20.

[0057] In one embodiment, the water meter reading device R1 can be Figure 5 In one embodiment, the water meter reading device 50 may include an image acquisition device. The water meter reading device R2 may be... Figure 1 In one embodiment, the water meter reading device 10 does not include an image acquisition device. Therefore, the verification result of the verification procedure performed by the water meter reading device R1 can be provided to other water meter reading devices R2 to RM in the same area as a reference. For example...

[0058] In detail, when the water meter reading displayed by water meter reader R1 is much greater than the water usage reading determined by water meter reader R1 based on the audio signal, it means that the collected audio signal may be affected by environmental noise or other external factors, causing water meter reader R1 to be unable to accurately estimate the number of impeller rotations based on the audio signal. In this case, water meter reader R1 can notify another water meter reader R2 located in the same area through the warning message Msg1, so that the other water meter reader R2 can take corresponding actions, such as correcting the water usage reading. Alternatively, water meter reader R1 can send the warning message Msg1 to server device 20 via network repeater H1, and server device 20 can uniformly correct or otherwise process the water usage readings reported by water meter readers R1 to RM.

[0059] In summary, in the embodiments of the present invention, the water usage reading of the water meter can be determined based on the audio signal generated by the rotation of the impeller inside the water meter. Compared to using an image acquisition device to achieve automatic water meter reading, the audio receiving device is cheaper and less prone to damage due to environmental factors. Furthermore, in the embodiments of the present invention, the accuracy and reliability of automatic water meter reading can be improved through mutual verification of the two automatic reading methods, and any abnormal phenomena in the water meter can be detected in real time.

[0060] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A water meter reading device, suitable for installation on a water meter, characterized in that, include: A radio receiver used to collect audio signals; The processor, coupled to the radio receiving device, converts the audio signal into a spectrogram. By inputting the spectrogram into a neural network model, it detects multiple target feature blocks in the spectrogram and counts the number of these target feature blocks to determine the water usage of the water meter. The horizontal axis of the spectrogram represents time, the vertical axis represents frequency, and the spectrogram includes M*N feature points, where M and N are positive integers. The number of target feature blocks reflects the number of rotations of the impeller of the water meter. The radio receiving device acquires the audio signal within a preset time interval. The processor inputs the spectrum corresponding to the preset time interval into the neural network model to determine the number of rotations of the impeller within the preset time interval.

2. The water meter reading device according to claim 1, characterized in that, Also includes: A transmission device, coupled to the processor, transmits the reading of the water meter to the server device.

3. The water meter reading device according to claim 1, characterized in that, Before converting the audio signal into the spectrogram, the processor performs noise reduction processing on the audio signal.

4. The water meter reading device according to claim 1, characterized in that, The processor uses Fast Fourier Transform to convert the audio signal into the spectrogram.

5. The water meter reading device according to claim 1, characterized in that, The neural network model includes a convolutional neural network model.

6. The water meter reading device according to claim 1, characterized in that, It also includes an image acquisition device to capture a degree display area of ​​the water meter to acquire a degree image. The processor acquires the displayed degree of the water meter based on the degree image and performs a verification procedure based on the displayed degree and the water usage. If the displayed degree and the water usage fail the verification, the processor issues a warning message through a transmission device.

7. The water meter reading device according to claim 6, characterized in that, If the difference between the displayed temperature and the water usage is greater than a threshold, the processor determines that the display temperature and the water usage have failed verification; if the difference between the displayed temperature and the water usage is not greater than the threshold, the processor determines that the display temperature and the water usage have passed verification.

8. A water meter reading method, applicable to a water meter reading device installed on a water meter, characterized in that, The method includes: Using a radio to collect audio signals; Convert the audio signal into a spectrum; Multiple target feature blocks in the spectrogram are detected by inputting the spectrogram into a neural network model; and The water consumption of the water meter is determined by counting the number of target feature blocks, wherein the horizontal axis of the spectrum represents time, the vertical axis of the spectrum represents frequency, and the spectrum includes M*N feature points, where M and N are positive integers. The number of target feature blocks reflects the number of rotations of the impeller of the water meter. Detecting the multiple target feature blocks in the spectrum by inputting the spectrum into the neural network model includes: inputting a spectrum corresponding to a preset time interval into the neural network model to determine the number of rotations of the impeller within the preset time interval.

9. The water meter reading method according to claim 8, characterized in that, The method further includes: using a transmission device to transmit the reading of the water meter to a server device.

10. The water meter reading method according to claim 8, characterized in that, The method further includes: performing noise reduction processing on the audio signal.

11. The water meter reading method according to claim 8, characterized in that, The step of converting the audio signal into the spectrogram includes: using a Fast Fourier Transform to convert the audio signal into the spectrogram.

12. The water meter reading method according to claim 8, characterized in that, The neural network model includes a convolutional neural network model.

13. The water meter reading method according to claim 8, characterized in that, The method further includes: The water meter's reading display area is photographed using an image acquisition device to obtain a reading image; The water meter's displayed reading is obtained based on the reading image, and a verification procedure is performed based on the displayed reading and the water usage reading; and If the displayed temperature and the water usage temperature fail to pass verification, a warning message will be issued via the transmission device.

14. The water meter reading method according to claim 13, characterized in that, If the difference between the displayed temperature and the water usage is greater than a threshold value, it is determined that the displayed temperature and the water usage have failed the verification; if the difference between the displayed temperature and the water usage is not greater than the threshold value, it is determined that the displayed temperature and the water usage have passed the verification.

Images