System and method for monitoring drainage and drain pipes for water pollution sources

Abstract

The present invention relates to the technical field of signal processing, and more particularly to a system and method for monitoring drains and drain pipes for water pollution sources. [Solution] The method includes step S1 of setting up drainage pipe monitoring points based on a drainage pipe network, step S2 of filtering the first audio information to obtain second audio information, step S3 of converting a spectrogram corresponding to a waveform segment of the second audio information into a two-dimensional image, step S4 of inputting the two-dimensional image into a noise model to determine whether the two-dimensional image is a noise image, and if it is a noise image and the adjacent two-dimensional image is not a noise image, generating a second sample image based on the noise waveform and the water leakage audio waveform corresponding to the two-dimensional image to train the leakage audio model, and step S5 of inputting the adjacent two-dimensional images into the leakage audio model to obtain an output result, and determining whether the pipe corresponding to the monitoring point is leaking based on the output result.

Description

[Technical Field]

[0001] The present invention relates to the technical field of signal processing, and in particular to a system and method for monitoring drains and drainpipes for water pollution sources. [Background technology]

[0002] Sewage and drainage pipe monitoring technology has been widely applied in urban drainage system management. Governments and drainage-related departments at all levels use this technology to regularly inspect and evaluate drainage pipes, timely identify and resolve pipe problems, and ensure the normal operation of drainage systems. In particular, comprehensive inspections of industrial park road drainage pipe networks have been conducted to detect pipe defects and establish quality problem lists for the park's road drainage pipe network. A similar prior art example is the Chinese patent publication number CN105464614A, which proposes a water supply and drainage pipe leak detection system and method, including a composite pipe and a leak detection device, with the leak detection device equipped with a detection circuit. The leak detection device has conductive layers connected to both poles, and a grounding electrode installed in a medium outside the composite pipe. When the entire composite pipe wall or the outer insulating layer is damaged, the insulation resistance of the leak detection device changes. This allows for early prediction of water supply and drainage pipe damage and accurate location of the damage, significantly improving preventative measures and emergency repair efficiency. Although the above patent documents have solved the problem of detecting pipe leakage, they suffer from low accuracy, inconvenient operation, and low efficiency. Summary of the Invention [Means for solving the problem]

[0003] In order to further solve the above technical problems, the present invention provides a method for monitoring drainage and drainage pipes of water pollution sources, the method comprising: Step S1: detecting the water quality of the target industrial park, obtaining the water pollution source and the drainage pipe network of the water pollution source, and setting up the drainage pipe monitoring point according to the drainage pipe network; Step S2: periodically acquiring first audio information of the drainage pipe corresponding to the monitoring point through a detection unit, and filtering the first audio information to obtain second audio information, where the second audio information is a continuous time series; a step S3 of dividing the waveform corresponding to the second audio information into a plurality of waveform segments of time T, and forming spectrograms corresponding to the waveform segments into two-dimensional images; a step S4 of determining whether the two-dimensional image is a noise image, and if the two-dimensional image is a noise image and the two-dimensional image adjacent to the two-dimensional image is not the noise image, obtaining a noise waveform of a first sample image that corresponds to the two-dimensional image and has the highest similarity, and generating a second sample image of a leakage sound model according to the noise waveform and each water leakage sound waveform, to train the leakage sound model; and step 5 of acquiring the adjacent two-dimensional image and inputting it into the leakage sound model, obtaining an output result, and determining whether the pipe corresponding to the monitoring point is leaking based on the output result.

[0004] In a preferred embodiment of the present invention, step S2 Step S21: collecting the first audio information of the drainage pipe corresponding to the monitoring point in real time through an audio collection module of the detection unit; Step S22: storing the first audio information in a storage unit, and performing Fourier transform on the first audio information to obtain a first frequency range of all audio signals in the first audio information, and determining a plurality of frequencies in the first frequency range whose energy is greater than a preset value and whose duration is less than a predetermined time as target frequencies; and step S23 of filtering out the noise corresponding to the target frequency from within the first frequency range via a filtering unit to obtain the second audio information.

[0005] In a preferred embodiment of the present invention, step S3 The method includes obtaining a corresponding waveform based on the second audio information through an audio conversion circuit, dividing the waveform into a plurality of waveform segments of time T with different starting points, and forming a spectrogram corresponding to each of the waveform segments into the two-dimensional image.

[0006] In a preferred embodiment of the present invention, step S4 step S41 of inputting each of the two-dimensional images into the noise model, outputting the similarity between the two-dimensional image and each of the first sample images in the noise model, setting the similarity with the largest value as a maximum similarity, and if the maximum similarity is greater than a set threshold, determining that the two-dimensional image is a noise image, or if not, determining that the two-dimensional image is not a noise image; If the two-dimensional image is considered to be a noise image and the two-dimensional image adjacent to the two-dimensional image is considered not to be the noise image, a step S42 is performed in which the first sample image corresponding to the maximum similarity is taken as a target image, and a noise set type and the noise set time series corresponding to the target image are obtained based on the target image. and step S43 of acquiring waveforms of the noise set, combining the waveforms of the noise set and any number of waveforms of water leakage sounds in different time series to acquire multiple mixed sounds, respectively dividing the waveforms corresponding to each of the mixed sounds into multiple waveform segments with a period T, respectively acquiring spectrograms corresponding to each of the waveform segments, respectively using the spectrograms of all the waveform segments corresponding to the mixed sounds as second sample images, and training the leakage sound model based on the second sample images.

[0007] In a preferred embodiment of the present invention, the noise model is constructed and trained by: A plurality of noise signals in the target industrial park is obtained from a storage unit, and a plurality of noise set waveforms are obtained by randomly combining any number of the noise signals in different time series, and each noise set waveform is divided into a plurality of noise waveform segments of time T, and the spectrogram of each of the noise waveform segments and the noise are synthesized into a first sample image, and all of the first sample images are used as learning data to train the noise model.

[0008] In a preferred embodiment of the present invention, step S5 The method includes acquiring the adjacent two-dimensional images and inputting them into the leakage sound model, comparing the adjacent two-dimensional images with each of the second sample images to acquire a first maximum similarity between the adjacent two-dimensional images and the second sample images, and if the first maximum similarity is greater than a first set value, determining that the pipe at the monitoring point is leaking, and acquiring a leak type corresponding to the water leakage sound by the second sample image corresponding to the first maximum similarity.

[0009] In a preferred embodiment of the present invention, step S5 If the first maximum similarity value is smaller than the first set value and equal to or greater than a second set value, the two-dimensional image after the adjacent two-dimensional image is set as a definitive image, and the definitive image is input into the leakage sound model; step S5 is repeated to obtain a second maximum similarity value corresponding to the definitive image; if the second maximum similarity value is larger than the second set value and both the second sample image corresponding to the first maximum similarity value and the second sample image corresponding to the second maximum similarity value contain the same water leakage sound, it is determined that a leak has occurred from the piping at the monitoring point and that the leak type is the leak type corresponding to the water leakage sound; and if at least one of the first maximum similarity value or the second maximum similarity value is less than the second set value, it is determined that no leak has occurred in the piping corresponding to the monitoring point.

[0010] In a preferred embodiment of the present invention, the distance between the monitoring point and the drain pipe is shorter than a set distance.

[0011] In a preferred embodiment of the present invention, there is provided a monitoring system for drainage and drain pipes of a water pollution source for carrying out the method, the system comprising: An installation unit for detecting the water quality of the target industrial park, obtaining the water pollution source and the drainage pipe network of the water pollution source, and installing the monitoring point of the drainage pipe according to the drainage pipe network; a monitoring unit that periodically acquires first audio information of the drainage pipe corresponding to the monitoring point through a detection unit, and filters the first audio information to acquire second audio information, where the second audio information is a continuous time series; a division unit for dividing a waveform corresponding to the second audio information into a plurality of waveform segments of time T, and forming a spectrogram corresponding to each of the waveform segments into a two-dimensional image; a determining unit for determining whether the two-dimensional image is a noise image; a model training unit for obtaining a noise waveform of a first sample image having the highest similarity corresponding to the two-dimensional image when the two-dimensional image is a noise image and the two-dimensional image adjacent to the two-dimensional image is not the noise image, and generating a second sample image of a leakage voice model according to the noise waveform and each water leakage voice waveform, thereby training the leakage voice model; and a judgment unit that acquires the adjacent two-dimensional images, inputs them into the leakage sound model, obtains an output result, and judges whether the pipe corresponding to the monitoring point is leaking based on the output result.

[0012] Compared with the prior art, the present invention has at least the following beneficial effects: The present invention provides a method for detecting whether the second audio information is noise based on the two-dimensional image and a noise model, and a method for detecting whether the second audio information is noise based on the two-dimensional image and a noise model. ... By acquiring a target image and a noise set, the waveform of each noise type is acquired according to the noise type in the noise set, the waveform of the noise set is acquired according to the noise set time series, a second sample image is acquired, and the leaky voice model is trained based on the second sample image, the adjacent two-dimensional image is acquired and input into the leaky voice model, and the second sample image that is most similar to the adjacent two-dimensional image is output. If the first maximum similarity value corresponding to the second sample image is greater than the first set value, it is determined that a leak has occurred. If the first maximum similarity value is less than the first set value and greater than or equal to the second set value, a confirmed image is input into the leaky voice model to determine whether the pipe at the monitoring point has a leak. Through the mutual cooperation of the above technical solutions, pipe leaks can be accurately identified. [Brief explanation of the drawings]

[0013] [Figure 1] 1 is a flowchart of a method for monitoring drainage and drainage pipes of water pollution sources according to the present invention. [Figure 2] 1 is a block diagram of a monitoring system for drainage and drainage pipes of water pollution sources according to the present invention; DETAILED DESCRIPTION OF THE INVENTION

[0014] In order to clarify the objectives, claims and advantages of the present invention, the present invention will be described in more detail below in conjunction with drawings and examples. It should be understood that the specific embodiments described in this specification are only used to explain the present invention, and are not used to limit the present invention.

[0015] The present invention provides a method for monitoring drainage and drainage pipes of water pollution sources, as shown in FIG. 1, the method includes the following steps: Step S1: Detect the water quality of the target industrial park, obtain the water pollution source and the drainage pipe network of the water pollution source, and set up the drainage pipe monitoring points according to the drainage pipe network.

[0016] Specifically, by installing the monitoring point and ensuring that the distance between the monitoring point and the piping is smaller than a predetermined distance, it is possible to ensure that the monitoring unit acquires high-quality first audio information, and further improve the accuracy of piping leak identification.

[0017] Step S2: periodically obtain first audio information of the drainage pipe corresponding to the monitoring point through a detection unit, and filter the first audio information to obtain second audio information, where the second audio information is a continuous time series.

[0018] Specifically, the audio collection module acquires the first audio information of the drainage pipe corresponding to the monitoring point in real time, and then obtains the target frequency range corresponding to the first audio information through the Fourier transform. The filtering unit filters out the frequency corresponding to the accidentally appearing audio, and obtains the second audio information, i.e., the unmeasurable noise of the accidental burst. The technical solution can filter out the burst noise, and further determine the audio set of the second audio information.

[0019] Step S3: Divide the waveform corresponding to the second audio information into a plurality of waveform segments of time T, and create a spectrogram corresponding to each waveform segment into a two-dimensional image.

[0020] Specifically, the second audio information is a filtered audio signal, and is converted into a corresponding waveform through an audio conversion circuit. Since the time starting point of the waveform corresponding to the first sample image in the noise model is unknown, in order to more accurately match the first sample image, the waveform is divided into multiple waveform segments with a time length T at different time starting points, and the corresponding spectrograms of the waveform segments are the two-dimensional images. Whether the second audio information is noise is determined based on the two-dimensional image and the noise model. This technical solution can more accurately identify whether the second audio information is noise.

[0021] Step S4: Determine whether the two-dimensional image is a noise image. If the two-dimensional image is a noise image and the two-dimensional image adjacent to the two-dimensional image is not the noise image, obtain the noise waveform of the first sample image that corresponds to the two-dimensional image and has the highest similarity, and generate a second sample image of the leakage voice model according to the noise waveform and each water leakage voice waveform, and train the leakage voice model.

[0022] Specifically, the noise model can determine the similarity between the input spectrogram segment and each sample image. Each of the two-dimensional images is input into the noise model and compared with each part of the sample image to obtain the target sample image that is most similar to the two-dimensional image. If the similarity corresponding to the target sample image is greater than the set threshold, the two-dimensional image is deemed to be a noise image; otherwise, the two-dimensional image is deemed not to be a noise image. In order to determine whether the pipe at the corresponding monitoring point is leaking using the adjacent two-dimensional images, the target image corresponding to the two-dimensional image is obtained, and the noise type in the noise set corresponding to the target image and the noise set time series are obtained. The waveform of each noise type is obtained based on the noise type in the noise set, and the waveform of the noise set is obtained based on the noise set time series. Therefore, by identifying a noise set in a noise time period and superimposing the water leakage sound in a different time series based on the noise set, the waveform period of the mixed sound after superposition is long, so that once a leak occurs in the piping network at the monitoring position during drainage, it can be quickly identified in a short time. The waveform of the mixed sound is divided into multiple waveform segments with a period T, and the spectrogram corresponding to each waveform segment is used as a second sample image. The leakage sound model is trained based on the second sample image. This technical solution can obtain an accurate leakage sound model and lay a foundation for quickly determining whether the piping network at the monitoring point is leaking.

[0023] Step 5: Obtain the adjacent two-dimensional image and input it into the leakage sound model, obtain the output result, and determine whether the pipe corresponding to the monitoring point is leaking based on the output result.

[0024] Specifically, the adjacent two-dimensional images are acquired and input into the leakage sound model, the adjacent two-dimensional images are compared with each of the second sample images by the leakage sound model, and the second sample image having the highest similarity to the adjacent two-dimensional image is output. When the first maximum similarity value corresponding to the second sample image is greater than the first set value, it is determined that the pipe corresponding to the monitoring point has leaked. When the first maximum similarity value is smaller than the first set value and equal to or greater than the second set value, it is explained that the adjacent two-dimensional images cannot accurately identify the water leakage sound. When sewage is discharged, the leak continues for a certain period of time and the leak type changes rapidly. Therefore, it is possible to re-determine whether a leak has occurred by using the 2D image after the adjacent 2D image, i.e., the final image, and the second maximum similarity value corresponding to the final image is greater than the second set value, and the water leakage sound contained in the mixed sound corresponding to the corresponding second sample image and the water leakage sound in the mixed sound corresponding to the second sample image corresponding to the adjacent image can be simultaneously determined, and it is deemed that the pipe corresponding to the monitoring point is leaking and the leakage type is the leakage type corresponding to the water leakage sound, and the technical solution can further improve the accuracy of pipe leakage identification.

[0025] Furthermore, step S2 Step S21: collecting the first audio information of the drainage pipe corresponding to the monitoring point in real time through an audio collection module of the detection unit; Step S22: storing the first audio information in a storage unit, and performing Fourier transform on the first audio information to obtain a first frequency range of all audio signals in the first audio information, and determining a plurality of frequencies in the first frequency range whose energy is greater than a preset value and whose duration is less than a predetermined time as target frequencies; and step S23 of filtering out the noise corresponding to the target frequency from within the first frequency range via a filtering unit to obtain the second audio information.

[0026] Specifically, the audio acquisition module acquires first audio information of the drainage pipe corresponding to the monitoring point in real time, stores the first audio information in the storage unit, and then performs Fourier transform to obtain the target frequency range corresponding to the first audio information. In industrial parks, there are frequent and measurable noises, but there are also occasional and relatively short-duration noises such as horns. Therefore, the filtering unit needs to filter out the frequencies corresponding to the randomly generated audio to obtain the second audio information, i.e., random bursts of unmeasurable noise. This technical solution is useful for filtering out burst noise to further determine the audio set of the second audio information.

[0027] Furthermore, step S3 A corresponding waveform is obtained based on the second audio information through an audio conversion circuit, and the waveform is divided into a plurality of waveform segments with different starting points and time T, and a spectrogram corresponding to each of the waveform segments is used as the two-dimensional image.

[0028] Specifically, the second audio information is a filtered audio signal, and is converted into a corresponding waveform through an audio conversion circuit. Since the time starting point of the corresponding waveform of the first sample image in the noise model is unknown, in order to more accurately match the first sample image, the waveform is divided into multiple waveform segments with a time length T at different time starting points, and the corresponding spectrograms of the waveform segments are used as the two-dimensional image. Whether the second audio information is noise is determined based on the two-dimensional image and the noise model. This technical solution can more accurately identify whether the second audio information is noise.

[0029] Furthermore, step S4 includes the following steps: Step S41: input each of the two-dimensional images into the noise model, and output the similarity between the two-dimensional image and each of the first sample images in the noise model; the similarity with the largest value is set as the maximum similarity; if the maximum similarity is greater than a set threshold, the two-dimensional image is deemed to be a noise image; if not, the two-dimensional image is deemed not to be a noise image; Specifically, the noise model is a model trained on spectrogram segments of a noise set that combines a number of commonly occurring and measurable simple noises and complex noises in the industrial park. The noise model determines the similarity between the input spectrogram segments and each sample image. Each of the 2D images is input into the noise model and compared with each part of the sample image to determine the target sample image that is most similar to the 2D image. If the similarity corresponding to the target sample image is greater than the set threshold, the 2D image is considered to be a noise image; otherwise, the 2D image is not considered to be a noise image. The technical solution can accurately determine whether the 2D image is a noise image, laying a foundation for more accurate determination of pipe leakage.

[0030] Step S42: if the two-dimensional image is deemed to be a noise image and the two-dimensional image adjacent to the two-dimensional image is deemed not to be the noise image, the first sample image corresponding to the maximum similarity is taken as a target image, and the noise set type and the noise set time series corresponding to the target image are obtained based on the target image.

[0031] Step S43: obtain the waveforms of the noise set, combine the waveforms of the noise set and any number of water leakage sound waveforms in different time series to obtain multiple mixed sounds, divide the waveforms corresponding to each of the mixed sounds into multiple waveform segments with a period T, obtain spectrograms corresponding to each of the waveform segments, and use the spectrograms of all the waveform segments corresponding to the mixed sounds as second sample images. Train the leakage sound model based on the second sample images.

[0032] Specifically, because the noise is related to manufacturing and production in industrial parks, the duration of the noise is long compared to the duration of sewage discharge, and the noise and water leakage sounds may overlap in some time periods. If the 2D image is a noise image and the adjacent 2D image after the 2D image is not a noise image, the adjacent 2D image is likely to be normal drainage sounds or sewage leakage sounds superimposed on the noise corresponding to the 2D image. Therefore, to determine whether the piping network at the corresponding monitoring point is leaking using the adjacent 2D images, the target image corresponding to the 2D image is obtained, the noise type and the noise set time series in the noise set corresponding to the target image are obtained, the waveform of each noise type is obtained based on the noise type in the noise set, and the waveform of the noise set is obtained based on the noise set time series. Therefore, by identifying the noise combination during the noise time period and superimposing the water leakage sound in different time series based on the noise combination, the waveform period of the mixed sound after superimposition is long, so that once a leak occurs in the piping network at the monitoring point during drainage, it can be quickly identified in a short time. The waveform of the mixed sound is divided into multiple waveform segments with a period T, and the spectrogram corresponding to each waveform segment is used as a second sample image. The leakage sound model is trained based on the second sample image. This technical solution can obtain an accurate leakage sound model and lay the foundation for quickly determining whether the piping network at the monitoring point is leaking.

[0033] Furthermore, the noise model is constructed and trained by: A plurality of noise signals in the target industrial park is obtained from a storage unit, and a plurality of noise set waveforms are obtained by randomly combining any number of the noise signals in different time series, and each noise set waveform is divided into a plurality of noise waveform segments of time T, and the spectrogram of each of the noise waveform segments and the noise are synthesized into a first sample image, and all of the first sample images are used as learning data to train the noise model.

[0034] Specifically, by reading multiple types of noise signals from the target industrial park from the storage unit, the noise may exist alone or multiple types may exist simultaneously, and there are multiple combinations of noise waveform time series. Therefore, to obtain relatively complete training data for the noise model, any number of the noise signals are combined with different time series to obtain multiple corresponding noise set waveforms. Furthermore, since the duration of the noise set's appearance is uncertain, the noise set waveform is divided into multiple noise waveform segments of time T, and the spectrograms corresponding to each noise waveform segment are used as the first sample images to train the noise model. Here, different noise waveform segments correspond to different frequency ranges or have different energies corresponding to the same frequency. This technical solution enables the acquisition of a highly accurate noise model and further improves the accuracy and speed of noise recognition.

[0035] Furthermore, step S5 The method includes acquiring the adjacent two-dimensional images and inputting them into the leakage sound model, comparing the adjacent two-dimensional images with each of the second sample images to acquire a first maximum similarity between the adjacent two-dimensional images and the second sample images, and if the first maximum similarity is greater than a first set value, determining that the pipe at the monitoring point is leaking, and acquiring a leak type corresponding to the water leakage sound by the second sample image corresponding to the first maximum similarity.

[0036] Specifically, the adjacent two-dimensional images are obtained and input into the leakage sound model, the adjacent two-dimensional images are compared with each of the second sample images using the leakage sound model, and the second sample image that is most similar to the adjacent two-dimensional image is output. If the first maximum similarity value corresponding to the second sample image is greater than the first set value, for example, 90%, it is determined that a leak has occurred in the pipe corresponding to the monitoring point, and the leak type is obtained based on the water leakage sound type in the mixed sound corresponding to the second sample image. The technical solution can accurately identify whether a leak has occurred in the pipe at the monitoring point.

[0037] Furthermore, step S5 If the first maximum similarity value is smaller than the first set value and equal to or greater than a second set value, the two-dimensional image after the adjacent two-dimensional image is set as a definitive image, and the definitive image is input into the leakage sound model; step S5 is repeated to obtain a second maximum similarity value corresponding to the definitive image; if the second maximum similarity value is larger than the second set value and both the second sample image corresponding to the first maximum similarity value and the second sample image corresponding to the second maximum similarity value contain the same water leakage sound, it is determined that a leak has occurred from the piping at the monitoring point and that the leak type is the leak type corresponding to the water leakage sound; and if at least one of the first maximum similarity value or the second maximum similarity value is less than the second set value, it is determined that no leak has occurred in the piping corresponding to the monitoring point.

[0038] Specifically, if the first maximum similarity value is smaller than the first set value and greater than or equal to the second set value, the adjacent two-dimensional image cannot accurately identify the water leakage sound. When sewage is discharged, it continues for a certain period of time and the leakage type does not change rapidly, so whether a leak has occurred can be re-determined by the two-dimensional image after the adjacent two-dimensional image, i.e., the final image. If the second maximum similarity value corresponding to the final image is greater than the second set value, the water leakage sound contained in the mixed sound corresponding to the corresponding second sample image and the water leakage sound in the mixed sound corresponding to the second sample image corresponding to the adjacent image can be simultaneously determined. It is determined that the pipe corresponding to the monitoring point is leaking and the leakage type is the leakage type corresponding to the water leakage sound. The technical solution can further improve the accuracy of pipe leakage identification.

[0039] Furthermore, the distance between the monitoring point and the drain pipe is less than a set distance.

[0040] Specifically, the above technical solution can ensure that the monitoring unit obtains high-quality first audio information, thereby improving the accuracy of identifying pipe leakage.

[0041] The present invention provides a monitoring system for drainage and drainage pipes of water pollution sources for carrying out the method, as shown in FIG. 2, the system includes: An installation unit for detecting the water quality of the target industrial park, obtaining the water pollution source and the drainage pipe network of the water pollution source, and installing the monitoring point of the drainage pipe according to the drainage pipe network; a monitoring unit that periodically acquires first audio information of the drainage pipe corresponding to the monitoring point through a detection unit, and filters the first audio information to acquire second audio information, where the second audio information is a continuous time series; a division unit for dividing a waveform corresponding to the second audio information into a plurality of waveform segments of time T, and forming a spectrogram corresponding to each of the waveform segments into a two-dimensional image; a determining unit for determining whether the two-dimensional image is a noise image; a model training unit for obtaining a noise waveform of a first sample image having the highest similarity corresponding to the two-dimensional image when the two-dimensional image is a noise image and the two-dimensional image adjacent to the two-dimensional image is not the noise image, and generating a second sample image of a leakage voice model according to the noise waveform and each water leakage voice waveform, thereby training the leakage voice model; and a judgment unit that acquires the adjacent two-dimensional images, inputs them into the leakage sound model, obtains an output result, and judges whether the pipe corresponding to the monitoring point is leaking based on the output result.

[0042] The technical features of the above-described embodiments can be combined in any combination for the sake of simplicity, and although not all possible combinations of the technical features in the above-described embodiments are described, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope described in this specification.

[0043] The above-described embodiments only represent some embodiments of the present invention, and although the description is more specific and detailed, it cannot be understood as limiting the patent scope of the present invention. It should be noted that those skilled in the art who fall within the protection scope of the present invention can make some modifications and improvements without departing from the concept of the present invention. Therefore, the patent protection scope of the present invention should be governed by the appended claims.

[0044] The above are only preferred embodiments of the present invention, and do not need to limit the present invention. Any modifications, equivalent replacements and improvements made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. 1. A method for monitoring water pollution sources, including drainage and drain lines, the method comprising: Step S1: detecting the water quality of the target industrial park, obtaining the water pollution source and the drainage pipe network of the water pollution source, and setting up the drainage pipe monitoring point according to the drainage pipe network; Step S2: periodically acquiring first audio information of the drain pipe corresponding to the monitoring point through a detection unit, and filtering the first audio information to obtain second audio information, where the second audio information is a continuous time series; a step S3 of dividing the waveform corresponding to the second audio information into a plurality of waveform segments of time T, and forming spectrograms corresponding to the waveform segments into two-dimensional images; Step S4: inputting the two-dimensional image into a noise model to determine whether the two-dimensional image is a noise image; if the two-dimensional image is a noise image and the two-dimensional image adjacent to the two-dimensional image is not the noise image, obtaining a noise waveform of a first sample image having the highest corresponding similarity to the two-dimensional image; generating a second sample image of the leakage sound model based on the noise waveform and each water leakage sound waveform; and training the leakage sound model. A method for monitoring drainage and drainage pipes of water pollution sources, comprising step S5 of acquiring the adjacent two-dimensional images and inputting them into the leak sound model, obtaining output results, and determining whether the pipe corresponding to the monitoring point is leaking based on the output results.

2. Step S2 Step S21: collecting the first audio information of the drainage pipe corresponding to the monitoring point in real time through an audio collecting module of the detection unit; Step S22: storing the first audio information in a storage unit, and performing Fourier transform on the first audio information to obtain a first frequency range of all audio signals in the first audio information, and determining a plurality of frequencies in the first frequency range whose energy is greater than a preset value and whose duration is less than a predetermined time as target frequencies; 2. The method for monitoring drainage and drainage pipes for water pollution sources according to claim 1, further comprising: a step S23 of filtering out noise corresponding to the target frequency from within a first frequency range through a filtering unit to obtain the second audio information.

3. Step S3 The method for monitoring drainage and drainage pipes of water pollution sources as described in claim 1, further comprising: acquiring a corresponding waveform through an audio conversion circuit based on the second audio information; dividing the waveform into a plurality of waveform segments of time T at different starting points; and forming a spectrogram corresponding to each of the waveform segments into the two-dimensional image.

4. Step S4 step S41 of inputting each of the two-dimensional images into the noise model, outputting the similarity between the two-dimensional image in the noise model and each of the first sample images, setting the similarity with the largest value as the maximum similarity, and determining that the two-dimensional image is a noise image if the maximum similarity is greater than a set threshold value, or determining that the two-dimensional image is not a noise image if the result is No; If the two-dimensional image is considered to be a noise image and the two-dimensional image adjacent to the two-dimensional image is considered not to be the noise image, a step S42 is performed in which the first sample image corresponding to the maximum similarity is taken as a target image, and a noise set type and a noise set time series corresponding to the target image are obtained based on the target image. and (S43) acquiring a waveform of the noise model, combining the waveform of the noise model with any number of waveforms of water leakage sounds in different time series to acquire a plurality of mixed sounds, respectively dividing the waveform corresponding to each of the mixed sounds into a plurality of waveform segments with a period T, respectively acquiring spectrograms corresponding to each of the waveform segments, respectively obtaining the spectrograms of all the waveform segments corresponding to the mixed sounds as second sample images, and training the leakage sound model based on the second sample images.

5. The noise model is constructed and trained by:

2. The method for monitoring drainage and drainage pipes of water pollution sources according to claim 1, further comprising: obtaining a plurality of noise signals in the target industrial park from a storage unit; randomly combining any number of the noise signals in different time series to obtain a plurality of noise set waveforms; dividing each noise set waveform into a plurality of noise waveform segments of time T; synthesizing the spectrogram of each noise waveform segment and the noise signal into a first sample image; and using all the first sample images as learning data to train the noise model.

6. Step S5 2. The method for monitoring drainage and drainage pipes of water pollution sources, as described in claim 1, further comprising: acquiring the adjacent two-dimensional images and inputting them into the leakage sound model; comparing the adjacent two-dimensional images with each of the second sample images to obtain a first maximum similarity between the adjacent two-dimensional images and the second sample images; and if the first maximum similarity is greater than a first set value, determining that the pipe at the monitoring point is leaking; and obtaining a leak type corresponding to the water leakage sound using the second sample image corresponding to the first maximum similarity.

7. Step S5 7. The method for monitoring drainage and drainage pipes of water pollution sources according to claim 6, further comprising: if the first maximum similarity value is smaller than the first set value and equal to or greater than a second set value, determining the two-dimensional image after the adjacent two-dimensional image as a definitive image, and inputting the definitive image into the leakage sound model; repeating step S5 to obtain a second maximum similarity value corresponding to the definitive image; if the second maximum similarity value is greater than the second set value and both the second sample image corresponding to the first maximum similarity value and the second sample image corresponding to the second maximum similarity value contain the same water leakage sound, determining that a leak has occurred in the pipe at the monitoring point and that the leak type is the leak type corresponding to the water leakage sound; and if at least one of the first maximum similarity value or the second maximum similarity value is less than the second set value, determining that no leak has occurred in the pipe corresponding to the monitoring point.

8. 2. The method for monitoring drainage and drainage pipes of water pollution sources according to claim 1, wherein the distance between the monitoring point and the drainage pipe is less than a set distance.

9. A monitoring system for drainage and drain pipes of water pollution sources for carrying out the method according to any one of claims 1 to 8, comprising: An installation unit for detecting the water quality of the target industrial park, obtaining the water pollution source and the drainage pipe network of the water pollution source, and installing the monitoring point of the drainage pipe according to the drainage pipe network; a monitoring unit that periodically acquires first audio information of the drain pipe corresponding to the monitoring point through a detection unit, and filters the first audio information to acquire second audio information, where the second audio information is a continuous time series; a division unit for dividing a waveform corresponding to the second audio information into a plurality of waveform segments of time T, and forming a spectrogram corresponding to each of the waveform segments into a two-dimensional image; a determining unit for determining whether the two-dimensional image is a noise image; a model training unit for obtaining a noise waveform of a first sample image having the highest similarity corresponding to the two-dimensional image when the two-dimensional image is a noise image and the two-dimensional image adjacent to the two-dimensional image is not the noise image, and generating a second sample image of a leakage voice model according to the noise waveform and each water leakage voice waveform, thereby training the leakage voice model; a judgment unit that acquires the adjacent two-dimensional images, inputs them into the leak sound model, obtains output results, and judges whether the pipe corresponding to the monitoring point is leaking based on the output results.

Images