A harmful gas monitoring device and method for a smart livestock farm

Abstract

The application relates to the technical field of photoacoustic signal denoising, in particular to a harmful gas monitoring device and method for a smart livestock farm, the method comprising the following steps: collecting a to-be-detected gas in the livestock farm and obtaining a photoacoustic signal of the to-be-detected gas; calculating a feature of the photoacoustic signal affected by a noise superposition signal, so as to determine a penalty factor when the photoacoustic signal is subjected to denoising processing, and realize photoacoustic signal denoising; processing the denoised photoacoustic signal into a photoacoustic signal, so as to detect a harmful gas concentration through a gas sound wave detection module in the harmful gas monitoring device. The application aims to improve the denoising effect of the photoacoustic signal, and finally effectively improve the detection accuracy of the harmful gas monitoring device.

Description

Technical Field

[0001] This application relates to the field of photoacoustic signal denoising technology, specifically to a device and method for monitoring harmful gases in a smart livestock farm. Background Technology

[0002] In recent years, with the development of livestock farming towards large-scale, intensive, and intelligent operations, high-density indoor centralized feeding has become a common choice for farms. However, excessively high concentrations of harmful gases in livestock sheds not only pollute the air but also lead to a high incidence of livestock diseases, decreased production performance, and serious threats to the health of livestock. Healthy farming is a prerequisite for the sustainable development of the livestock industry. Therefore, intelligent air environment monitoring devices for livestock farming are gradually being applied to actual production, collecting real-time gas data in livestock sheds, accurately monitoring the concentration of harmful gases, and providing protection for the health of livestock.

[0003] Currently, the commonly used method for detecting harmful gases is photoacoustic spectroscopy, which uses the sound waves generated after a substance absorbs light energy to analyze and detect harmful gases. However, when this method is applied to livestock farms, it is easily affected by other sounds such as livestock calls and equipment operation, resulting in the photoacoustic signal obtained by photoacoustic spectroscopy containing livestock noise and causing photoacoustic signal overlap. When using the VMD (Variational Mode Decomposition) algorithm to decompose the noisy signal, it is difficult to determine the optimal penalty factor due to the complex and variable noise in the livestock house environment. A small penalty factor may lead to mode aliasing, while a large penalty factor may lose local information. The denoising effect of the photoacoustic signal directly affects the monitoring effect of harmful gases. Summary of the Invention

[0004] To address the aforementioned technical problems, this application provides a device and method for monitoring harmful gases in a smart livestock farm, the specific technical solution of which is as follows:

[0005] In a first aspect, one embodiment of this application provides a method for monitoring harmful gases in a smart livestock farm, the method comprising the following steps:

[0006] Collect the gas to be tested in a livestock farm and obtain its photoacoustic signal;

[0007] Based on the energy distribution difference of each high-energy region relative to all low-energy regions in the photoacoustic signal, and the energy distribution characteristics of the high-energy regions, the first feature of the photoacoustic signal affected by the noise overlap signal is calculated.

[0008] Based on the bandwidth discreteness and phase fluctuation discreteness of all frame signals in the photoacoustic signal, the second characteristic of the photoacoustic signal affected by noise overlap signal is calculated;

[0009] By combining the first and second characteristics of the influence of noise overlap signal on the photoacoustic signal, a penalty factor is determined when applying the variational mode decomposition algorithm to the photoacoustic signal, which is then used for photoacoustic signal noise reduction.

[0010] The noise-reduced photoacoustic signal is processed into a photoacoustic signal, which is then used to detect the concentration of harmful gases through the gas acoustic wave detection module in the harmful gas monitoring device.

[0011] Preferably, the method for determining the high-energy region and the low-energy region is as follows:

[0012] The collected photoacoustic signals are converted into Mel spectrograms;

[0013] The density peak clustering algorithm is used to calculate the local density at each coordinate point in each Mel-language spectrogram, which is denoted as the local energy density.

[0014] The Mel language spectrogram is divided into multiple regions using a region growing segmentation algorithm; the difference in local energy density between coordinate points is used as the distance metric during the region growing process.

[0015] The maximum energy in each region is used as the input to the maximum inter-class variance, and the output is an energy threshold. Regions with energy values ​​greater than the threshold are high-energy regions, and vice versa.

[0016] Preferably, the first feature is calculated as follows: the sum of the energy distribution differences between all high-energy regions and all low-energy regions is multiplied by the energy distribution characteristics of the high-energy regions, and the result is taken as the first feature.

[0017] Preferably, the energy distribution difference is determined by the average of the total density energy between each high-energy region and all low-energy regions; wherein the total density energy is the sum of the products of the energy at each coordinate point in the high / low energy region and the corresponding energy density.

[0018] Preferably, the calculation expression for the energy distribution characteristics of the high-energy region is: , This indicates the energy distribution characteristics of high-energy regions. This represents the area of ​​the high-energy region in the Mel spectrogram. This represents the total area of ​​the Mel language spectrogram. The standard deviation of the distance between high-energy regions. This indicates the number of high-energy regions.

[0019] Preferably, the calculation method for the second feature is as follows:

[0020] Calculate the bandwidth discreteness of all frames in the photoacoustic signal;

[0021] Calculate the average value of the phase fluctuation dispersion of all frame signals at the same frequency in the photoacoustic signal;

[0022] Calculate the average value of the phase fluctuation of all frames of the photoacoustic signal at all frequency points;

[0023] The product of the bandwidth discreteness, the average value of the phase fluctuation discreteness, and the average value of the phase fluctuation is taken as the second feature.

[0024] Preferably, the method for analyzing phase fluctuations is as follows:

[0025] Obtain the phase values ​​of all frame signals at the same frequency point p;

[0026] Calculate the ideal frequency of frequency point p; based on the integral relationship between frequency and phase, the ideal phase value of all frame signals at the same frequency point p can be calculated using the ideal frequency.

[0027] The absolute value of the difference between the phase value of each frame of signal at the same frequency point p and the ideal phase value is taken as the phase fluctuation value of each frame of signal at frequency point p.

[0028] Preferably, the discreteness is determined by the discrete coefficients.

[0029] Preferably, the penalty factor is calculated using the following formula: ;in, Indicates the penalty factor. This represents the initial penalty factor. , This represents the preset weighting coefficient. , These represent the first and second characteristics of the photoacoustic signal being affected by noise superimposed signals, respectively.

[0030] Secondly, another embodiment of this application also provides a harmful gas monitoring device for a smart livestock farm. The device includes a gas acquisition module, a sound wave acquisition module, a gas sound wave detection module, and an alarm module, wherein:

[0031] The gas acquisition module is used to collect the gas to be tested in the livestock farm and filter out dust and water vapor in the gas through the filter in the module.

[0032] The acoustic wave acquisition module is used to acquire the photoacoustic signal of the gas to be detected through light absorption, thermal expansion and photoacoustic effect, and to denoise the photoacoustic signal. This module realizes the above-mentioned method for monitoring harmful gases in a smart livestock farm.

[0033] The gas acoustic wave detection module is used to analyze the denoised photoacoustic signal and detect the concentration of harmful gases;

[0034] The alarm module is used to judge the monitoring results of the gas acoustic detection module and to sound an alarm when the harmful gas in the gas to be detected exceeds the standard.

[0035] This application has at least the following beneficial effects:

[0036] 1. This application uses the "high-low energy region difference" to quantify the superposition degree of impulse noise such as livestock and poultry calls with the intrinsic gas signal. It can distinguish whether the "local energy surge" is a gas absorption peak or transient interference caused by livestock and poultry calls, thereby suppressing the risk of misjudging the effective signal as noise and preserving the intrinsic gas peak structure.

[0037] 2. This application utilizes both bandwidth and phase fluctuation indicators to be equally sensitive to the spectral broadening and phase drift caused by continuous mechanical noise (such as fans and feeding lines), thus overcoming the deficiency of single energy indicators failing in continuous noise scenarios and achieving comprehensive perception of complex acoustic environments.

[0038] 3. This application maps "pulse-type" and "continuous-type" noise characteristics to a single penalty factor, achieving adaptive adjustment of VMD decomposition parameters without increasing manual intervention. This prevents mode aliasing and avoids smoothing out gas absorption peaks due to excessive penalty, thereby improving the noise reduction effect on photoacoustic signals and ultimately effectively improving the detection accuracy of harmful gas monitoring devices. Attached Figure Description

[0039] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 This is a flowchart illustrating a method for monitoring harmful gases in a smart livestock farm, as provided in one embodiment of this application. Detailed Implementation

[0041] Example 1

[0042] One embodiment of this application provides a harmful gas monitoring device for a smart livestock farm. The device includes a gas acquisition module, a sound wave acquisition module, a gas sound wave detection module, and an alarm module. The gas acquisition module is used to collect the gas to be detected in the livestock farm and filter out dust and water vapor in the gas through a filter in the module. The sound wave acquisition module is used to acquire the photoacoustic signal of the gas to be detected through light absorption, thermal expansion, and photoacoustic effect, and to denoise the photoacoustic signal. This module implements the method in Embodiment 2 below. The gas sound wave detection module is used to analyze the denoised photoacoustic signal and detect the concentration of harmful gases. The alarm module is used to judge the monitoring results of the gas sound wave detection module and issue an alarm sound when the harmful gas in the gas to be detected exceeds the standard.

[0043] Example 2

[0044] One embodiment of this application provides a method for monitoring harmful gases in a smart livestock farm, as detailed in the following document. Figure 1 The method includes the following steps:

[0045] Step 1: Collect the gas to be tested in the livestock farm and obtain its photoacoustic signal.

[0046] A harmful gas monitoring device for a smart livestock farm is installed 2 meters from the ventilation opening. The gas collection module collects the gas (gas molecules) to be detected inside the livestock shed. The collection time is 2 seconds, with a collection interval of 4 seconds. Since the air inside the livestock shed may contain dust and water vapor, the device has a PTFE filter with a pore size of 0.22 micrometers to filter out fine dust. Semiconductor refrigeration is used to condense and remove water vapor, resulting in gas with impurities removed. Next, a laser and modulation signal generator in the acoustic wave acquisition module generate modulated light that matches the absorption peak of the target gas, providing an energy excitation source for the photoacoustic effect. Then, the gas concentration signal of the gas to be detected is converted into a sound pressure signal (photoacoustic signal) through the photoacoustic effect. The photoacoustic cell is a cylindrical resonant photoacoustic cell with a resonant frequency of 4.98 kHz (an empirical value). The modulation light generation and photoacoustic effect are well-known technologies, and their specific implementation will not be elaborated further.

[0047] Step 2: Calculate the characteristics of the photoacoustic signal affected by the noise superimposed signal in order to determine the penalty factor when performing noise reduction processing on the photoacoustic signal, thereby achieving photoacoustic signal noise reduction.

[0048] S1: Based on the energy distribution difference of each high-energy region in the photoacoustic signal relative to all low-energy regions, and the energy distribution characteristics of the high-energy regions, calculate the first feature of the photoacoustic signal affected by the noise overlap signal.

[0049] The photoacoustic effect generates a photoacoustic signal with the same period as the modulation frequency. This signal has a very small amplitude, is symmetrical about the central peak, and is considered a weak signal. The modulation frequency is usually consistent with the resonant frequency of the photoacoustic cell. Because gas molecules selectively absorb light, only absorption occurs within a specific wavelength range. Typically, there is a functional relationship between the relative spectral energy distribution and wavelength in the spectrum. If the gas contains both hydrogen sulfide and ammonia, the resulting photoacoustic signal will contain two peaks (the peak values ​​are related to the gas concentration), corresponding to the two harmful gases respectively, which can be distinguished by wavelength. Therefore, the energy distribution at consistent wavelengths in a pure photoacoustic signal is usually relatively uniform. If livestock or poultry emit loud noise in the monitoring environment, the intensity of this sound signal is usually greater than the weak photoacoustic signal from the gas. This causes the collected photoacoustic signal to include environmental noise, resulting in signal overlap and uneven energy distribution. Specifically, the signal energy is higher at overlapping noise points, while the energy remains unchanged at non-overlapping points.

[0050] Specifically, this application converts the acquired photoacoustic signal into a Mel spectrogram. Frame-by-frame windowing is performed with a frame length of 10ms, a step size of 5ms, and a Hamming window function. Short-time Fourier transform is used for spectrum calculation. Density peak clustering algorithm is used to calculate the local density at each coordinate point in each Mel spectrogram, denoted as the local energy density. The difference in energy values ​​between coordinate points is used as the clustering distance between coordinate points in the density peak clustering algorithm.

[0051] Next, a region growing segmentation algorithm is used to divide the Mel language spectrogram into multiple regions. The difference in local energy density between each coordinate point is used as the distance metric in the region growing process, so that the energy in each region is basically equal.

[0052] The maximum energy in each region is used as the input to the maximum inter-class variance, and the output is an energy threshold. Regions with energy values ​​greater than the threshold may be high-energy regions containing overlapping noise, while regions with energy values ​​less than the threshold are low-energy regions without overlap. Due to the complexity and variability of noise signals, signal overlap is irregular, and high-energy regions in the spectrogram are also distributed in a blocky and discrete manner.

[0053] Among them, the generation of Mel language spectrograms, density peak clustering algorithm, and maximum inter-class variance are well-known techniques, and their specific implementations will not be elaborated here.

[0054] Based on the above analysis, the first characteristic of the photoacoustic signal being affected by the noise overlap signal is calculated, which is used to characterize the degree of overlap between the photoacoustic signal and the noise signal:

[0055]

[0056] in, The first characteristic indicating that photoacoustic signals are affected by noise superimposed signals is... This indicates the energy distribution characteristics of high-energy regions. Indicates the number of high-energy regions. This represents the m-th high-energy region. This represents the degree of energy distribution difference between the m-th high-energy region and all low-energy regions;

[0057] in, , This represents the area of ​​the high-energy region in the Mel language spectrogram, and its value is equal to the sum of the number of coordinate points within each high-energy region. This represents the total area of ​​the Mel spectrogram, and its value is equal to the number of coordinate points in the Mel spectrogram. This represents the standard deviation of the distance between high-energy regions. The distance between each region is equal to the Euclidean distance between the coordinates of the points with the highest local energy density within each high-energy region. The larger the area of ​​the high-energy region, the more dispersed its distribution, and the higher the density-energy difference. The larger the value, the higher the overlap between the photoacoustic signal and the noise signal. Euclidean distance is a well-known technique, and its specific implementation will not be elaborated upon. It should be noted that when... season ;when season .

[0058] in, , Indicates the number of low-energy regions. This represents the nth low-energy region. , Let represent the total energy density of the m-th high-energy region and the n-th low-energy region, respectively. Their values ​​are equal to the sum of the products of the energy at each coordinate point within that high / low energy region and the corresponding energy density. The larger the value, the greater the difference in energy distribution between the two regions; the higher the energy density in the high-energy region, the lower the energy density in the low-energy region, and the greater the density-energy difference. The larger the value, the greater the likelihood of signal overlap noise in that area. The larger the value, the better.

[0059] It should be noted that when When this occurs, it indicates that the entire Mel language spectrum belongs to a high-energy region, meaning there is only one high-energy region. Therefore, let... It is the sum of the products of the energy and the corresponding energy density of all coordinate points in the Mel language spectrogram.

[0060] S2: Based on the bandwidth discreteness and phase fluctuation discreteness of all frame signals in the photoacoustic signal, calculate the second characteristic of the photoacoustic signal affected by noise overlap signal.

[0061] Further analysis reveals that if photoacoustic signals overlap due to noise, the complexity of the noise can lead to variations in its bandwidth. This noise might increase energy at frequencies that were originally low or nonexistent in the photoacoustic signal, disrupting the stable bandwidth of the pure signal. For example, partial overlap between a noise signal and a peak in the photoacoustic signal could increase the energy near that peak, resulting in a larger bandwidth. This increased bandwidth can also cause frequency changes, leading to phase fluctuations. In other words, unstable bandwidth variations can cause phase changes to be disordered. Even if the frequency remains unchanged, signal overlap can alter the photoacoustic signal waveform, causing phase fluctuations.

[0062] Specifically, this application performs framing and windowing processing on the photoacoustic signal, with each frame being 20ms and windowed using a Hamming window. A Fourier transform is performed on each frame to obtain its amplitude and phase spectra. The power spectrum of each frame is obtained by squared the amplitude spectrum, and the 99% energy bandwidth within the power spectrum is calculated. The bandwidths of all frames of the photoacoustic signal are then combined to form a bandwidth sequence. ,in , , These represent the bandwidths of the photoacoustic signals in the 1st, 2nd, and Tth frames, respectively, where T represents the total number of frames.

[0063] To avoid phase entanglement due to excessive phase shift, an adjacent frame difference correction method is used to adjust the actual phase sequence of each frequency point in each frame. Perform untangling, where This represents the phase sequence at frequency point p. , , These represent the phase values ​​of frequency point p in the 1st, 2nd, and Tth frames, respectively.

[0064] The ideal frequency of each frequency point p is: Where N represents the number of sampling points in each frame of signal, taken as an empirical value of 512, and p represents the p-th sampling point. This represents the sampling frequency, the number of spectra collected per second. We take an empirical value of 200, from which the ideal frequency for each frequency point can be calculated.

[0065] At the ideal frequency, it is assumed that signals will not overlap. Based on the integral relationship between frequency and phase, the ideal phase value sequence at each frequency point can be calculated from the ideal frequency. ,in , , Let p represent the ideal phase values ​​at frequency point p in frames 1, 2, and T, respectively. The ideal phase value sequence... With actual phase sequence The absolute value of the difference is used as the phase fluctuation sequence. ,in , , These represent the phase fluctuation values ​​between the actual phase and the ideal phase at frequency point p in frames 1, 2, and T, respectively.

[0066] Based on the above analysis, a second characteristic of the photoacoustic signal affected by overlapping noise signals is calculated to characterize the fluctuation of the photoacoustic signal bandwidth and phase under the influence of overlapping noise signals:

[0067]

[0068] in, The second characteristic indicating that the photoacoustic signal is affected by noise superimposed signals, The coefficient of variation represents the discreteness of the bandwidth sequence, and its value is equal to the ratio of the standard deviation to the mean of the bandwidths of all frames in the bandwidth sequence. This represents the average of the phase fluctuation values ​​of all frames at all frequency points, which is the matrix distance between the actual phase matrix and the ideal phase matrix. The actual phase matrix is ​​constructed from the actual phase sequence of each frequency point across all frames, with dimensions of... The ideal phase matrix is ​​constructed using the same method as the actual phase matrix; specific details will not be elaborated further. This indicates the number of sampling points in each frame of the signal. The discrete coefficients of the phase fluctuation sequence at frequency point p are represented by their values ​​and are calculated using the same method as... The same applies, so I won't go into details here.

[0069] If the bandwidth of the photoacoustic signal varies more and more randomly between frames, and the phase fluctuation value is also more and more random, the value of the second characteristic value B is also larger, indicating that the photoacoustic signal is more severely affected by noise and overlapping signals.

[0070] S3: Combining the first and second features of the influence of noise overlap signal on the photoacoustic signal, determine the penalty factor when applying the variational mode decomposition algorithm to the photoacoustic signal, and use it for photoacoustic signal noise reduction.

[0071] When livestock farms use photoacoustic spectroscopy for hazardous gas monitoring, the collected photoacoustic signals contain a large amount of complex environmental noise from livestock and poultry, leading to signal overlap and directly affecting the monitoring accuracy. Existing VMD (Variational Mode Decomposition) algorithms struggle to determine the optimal penalty factor when decomposing noisy signals due to the complex and variable nature of livestock house noise. A small penalty factor may cause mode aliasing, while a large penalty factor may lose local information. Larger primary and secondary features indicate greater influence of overlapping noise signals on the photoacoustic signal, requiring a larger penalty factor to improve noise resistance. Conversely, smaller features indicate less noise, with the photoacoustic signal primarily containing gaseous signals; the detailed features of the signal are crucial for subsequent hazardous gas monitoring, necessitating a smaller penalty factor.

[0072] Based on the above analysis, the penalty factor is calculated, which plays a role in balancing the signal-to-noise ratio and decomposition accuracy during the signal decomposition process.

[0073]

[0074] in, Indicates the penalty factor. This represents the initial penalty factor, with a value ranging from 100 to 3000. The preferred initial value for this scheme is 400. Based on experience, it is taken as twice the sampling frequency, which is the value of this scheme. The minimum value is 400, and the maximum value is 3000. , The preset weighting coefficients are all taken as empirical values ​​of 0.5. If there is a lot of noise and severe signal overlap in the photoacoustic signal, the values ​​of the first feature A and the second feature B will also be relatively large, resulting in a penalty factor. The value is also relatively large, indicating that the noise resistance is relatively strong.

[0075] Punishment factor As input parameters for the VMD algorithm, the photoacoustic signal is decomposed into multiple IMF components, with the number of decomposition layers set to an empirical value of 10. Then, the wavelet threshold denoising algorithm is used to selectively retain and reconstruct the IMF components to obtain the denoised photoacoustic signal.

[0076] Step 3: The noise-reduced photoacoustic signal is passed through the gas acoustic wave detection module in the harmful gas monitoring device to detect the concentration of harmful gases.

[0077] The weak photoacoustic signal, after noise reduction, is amplified by a preamplifier within the hazardous gas monitoring device. Then, a lock-in amplifier demodulates the signal to form an electrical signal, also known as a photoacoustic signal, enhancing its intensity and characteristics. This photoacoustic signal is then input to the gas acoustic wave detection module to detect the concentration of hazardous gases. If the detected hazardous gas concentration exceeds the standard, the alarm module within the device sounds an alarm, alerting livestock farm staff to properly handle the hazardous gases and prevent excessive levels in the livestock sheds from harming livestock health and spreading diseases.

[0078] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not invented in this application.

[0079] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope.

Claims

1. A method for monitoring harmful gases in a smart livestock farm, characterized in that, The method includes the following steps: Collect the gas to be tested in a livestock farm and obtain its photoacoustic signal; Based on the energy distribution difference of each high-energy region relative to all low-energy regions in the photoacoustic signal, and the energy distribution characteristics of the high-energy regions, the first feature of the photoacoustic signal affected by the noise overlap signal is calculated. Based on the bandwidth discreteness and phase fluctuation discreteness of all frame signals in the photoacoustic signal, the second characteristic of the photoacoustic signal affected by noise overlap signal is calculated; By combining the first and second characteristics of the influence of noise overlap signal on the photoacoustic signal, a penalty factor is determined when applying the variational mode decomposition algorithm to the photoacoustic signal, which is then used for photoacoustic signal noise reduction. The noise-reduced photoacoustic signal is processed into a photoacoustic signal, which is then used to detect the concentration of harmful gases through the gas acoustic wave detection module in the harmful gas monitoring device.

2. The method for monitoring harmful gases in a smart livestock farm as described in claim 1, characterized in that, The method for determining the high-energy region and the low-energy region is as follows: The collected photoacoustic signals are converted into Mel spectrograms; The density peak clustering algorithm is used to calculate the local density at each coordinate point in each Mel-language spectrogram, which is denoted as the local energy density. The Mel language spectrogram is divided into multiple regions using a region growing segmentation algorithm; the difference in local energy density between coordinate points is used as the distance metric during the region growing process. The maximum energy in each region is used as the input to the maximum inter-class variance, and the output is an energy threshold. Regions with energy values ​​greater than the threshold are high-energy regions, and vice versa.

3. The method for monitoring harmful gases in a smart livestock farm as described in claim 1, characterized in that, The first feature is calculated as follows: the sum of the energy distribution differences between all high-energy regions and all low-energy regions is multiplied by the energy distribution characteristics of the high-energy regions, and the result is used as the first feature.

4. The method for monitoring harmful gases in a smart livestock farm as described in claim 3, characterized in that, The energy distribution difference is determined by the average of the total density energy between each high-energy region and all low-energy regions; wherein the total density energy is the sum of the products of the energy at each coordinate point in the high / low energy region and the corresponding energy density.

5. The method for monitoring harmful gases in a smart livestock farm as described in claim 3, characterized in that, The calculation expression for the energy distribution characteristics of the high-energy region is as follows: , This indicates the energy distribution characteristics of high-energy regions. This represents the area of ​​the high-energy region in the Mel spectrogram. This represents the total area of ​​the Melanographic spectrogram. The standard deviation of the distance between high-energy regions. This indicates the number of high-energy regions.

6. The method for monitoring harmful gases in a smart livestock farm as described in claim 1, characterized in that, The calculation method for the second feature is as follows: Calculate the bandwidth discreteness of all frames in the photoacoustic signal; Calculate the average value of the phase fluctuation dispersion of all frame signals at the same frequency in the photoacoustic signal; Calculate the average value of the phase fluctuation of all frames of the photoacoustic signal at all frequency points; The product of the bandwidth discreteness, the average value of the phase fluctuation discreteness, and the average value of the phase fluctuation is taken as the second feature.

7. The method for monitoring harmful gases in a smart livestock farm as described in claim 6, characterized in that, The analysis method for the phase fluctuation is as follows: Obtain the phase values ​​of all frame signals at the same frequency point p; Calculate the ideal frequency of frequency point p; based on the integral relationship between frequency and phase, the ideal phase value of all frame signals at the same frequency point p can be calculated using the ideal frequency. The absolute value of the difference between the phase value of each frame of signal at the same frequency point p and the ideal phase value is taken as the phase fluctuation value of each frame of signal at frequency point p.

8. The method for monitoring harmful gases in a smart livestock farm as described in claim 6, characterized in that, The discreteness is determined by the discrete coefficients.

9. The method for monitoring harmful gases in a smart livestock farm as described in claim 1, characterized in that, The formula for calculating the penalty factor is: ;in, Indicates the penalty factor. This represents the initial penalty factor. , This represents the preset weighting coefficient. , These represent the first and second characteristics of the photoacoustic signal being affected by noise superimposed signals, respectively.

10. A harmful gas monitoring device for a smart livestock farm, the device comprising a gas acquisition module, a sound wave acquisition module, a gas sound wave detection module, and an alarm module, characterized in that: The gas acquisition module is used to collect the gas to be tested in the livestock farm, and the filter in the module removes dust and water vapor from the gas. The acoustic wave acquisition module is used to acquire the photoacoustic signal of the gas to be detected through light absorption, thermal expansion and photoacoustic effect, and to denoise the photoacoustic signal. This module realizes the method for monitoring harmful gases in a smart livestock farm as described in any one of claims 1-9. The gas acoustic wave detection module is used to analyze the denoised photoacoustic signal and detect the concentration of harmful gases; The alarm module is used to judge the monitoring results of the gas acoustic wave detection module and to sound an alarm when the harmful gas in the gas to be detected exceeds the standard.

Images