An earthquake identification method based on urban surveillance camera sensor network

Abstract

The present invention discloses an earthquake recognition method based on a city monitoring camera sensor network, normalizes and filters earthquake audio and negative sample audio; writes the audio category and monitoring location information of earthquake audio and negative sample audio into a label file; extracts the frequency domain features of earthquake audio to obtain a Mel spectrum map; extracts the time domain features of earthquake audio to obtain a zero crossing rate map; fuses the Mel spectrum map and the zero crossing rate map to obtain a fusion feature; inputs the obtained fusion feature into a Transformer model to obtain a feature tensor; inputs the feature tensor into an improved LSTM model to obtain features on the earthquake audio time series; uses an outlier detection method to correct earthquake events misjudged by earthquake observation points; uses a spatial interpolation method to analyze earthquake events identified by multiple earthquake observation points to obtain earthquake spatiotemporal distribution and evolution information. The present invention gives full play to the observation advantages of the city monitoring sensor network, such as large number, high density, and fast transmission, and provides an opportunity for earthquake monitoring with high spatiotemporal resolution.

Description

Technical Field

[0001] The present invention relates to an earthquake recognition method based on a city monitoring camera sensor network, and belongs to the technical field of computer hearing and deep learning. Background Art

[0002] High temporal and spatial resolution earthquake data is crucial for earthquake disaster warning and post-disaster modeling. Earthquakes are characterized by suddenness, destructiveness, and difficulty in defense. Due to factors such as frequent human activities, active crustal movement, and changes in geological conditions, earthquakes have occurred frequently around the world in recent years. However, current earthquake monitoring methods are difficult to meet the requirements of high temporal and spatial resolution earthquake data required for large, comprehensive, and detailed urban earthquake modeling, which means that better monitoring methods are needed.

[0003] Today, earthquake monitoring methods mainly use seismic networks and high-altitude remote sensing technology. The seismic network monitors earthquakes and obtains earthquake data by deploying precision instruments such as seismometers, intensity meters, and intensity meters. However, the deployment cost of monitoring instruments is high, and it is difficult to deploy them over a large area and at a high density, resulting in low spatial resolution of the collected earthquake data. High-altitude remote sensing technology uses satellites to achieve large-scale observations and has high spatial resolution. However, the satellite revisit period is long, and it is impossible to monitor the occurrence of earthquakes, and the temporal resolution of the acquired earthquake data is low. Summary of the invention

[0004] The technical problem to be solved by the present invention is to overcome the defects of the prior art and provide an earthquake identification method based on a city monitoring camera sensor network. City monitoring cameras are widely distributed and can be used as earthquake observation points, which improves the spatial density of monitoring and the spatial resolution of earthquake data. At the same time, the city monitoring sensor network is based on 5G communication technology for all-weather observation, high-speed and rapid transmission of earthquake data, which improves the temporal resolution and enhances the city's earthquake monitoring and disaster warning capabilities.

[0005] Preferably, the present invention provides an earthquake identification method based on a city monitoring camera sensor network, comprising:

[0006] Obtain earthquake monitoring data collected from multiple earthquake observation points, including earthquake audio, negative sample audio, and monitoring location information;

[0007] Normalize and filter the earthquake audio and negative sample audio;

[0008] Write the audio category of the earthquake audio, the audio category of the negative sample audio, and the monitoring location information into the label file;

[0009] Extract the frequency domain features of the seismic audio and obtain the Mel-spectrogram.

[0010] Preferably, the time domain characteristics of the seismic audio are extracted to obtain a zero-crossing rate map.

[0011] Prioritize, the mel-spectrogram and the zero-crossing rate map are fused to obtain fused features.

[0012] Prioritize, the obtained fused features are input into the Transformer model to obtain feature tensors.

[0013] Prioritize, the feature tensor is input into the improved LSTM model to obtain the features on the earthquake audio time series and realize the recognition of earthquake audio.

[0014] Preferably, an outlier detection method is used to correct earthquake events misjudged by earthquake observation points, and a single surveillance camera can be used as a earthquake observation point.

[0015] Preferably, spatial interpolation is used to analyze earthquake events identified at multiple earthquake observation points to obtain information on the spatiotemporal distribution and evolution of earthquakes.

[0016] Firstly, the calculation formula of the mel-spectrogram is:

[0017] S Mel-Spec (t,f)=LogScale(MelFilterBank(FFT(Window(s(t)))))

[0018] Among them, S Mel-Spec (t,f) represents the Mel-spectrogram function, t is time, f is frequency, LogScale represents the logarithmic transform of the output of the Mel filter, MelFilterBank represents the result of FFT processing through the Mel filter bank, FFT represents the fast Fourier transform of the signal frame after the window function, and Window(s(t)) represents framing the signal s(t) and applying the window function to each frame.

[0019] Preferably, the zero-crossing rate map is calculated as:

[0020]

[0021] Where ZCR(b) represents the zero-crossing rate graph of the b-th frame, N represents the total number of signal values ​​in the b-th frame of the seismic audio, sign(s) represents the sign function, s represents the signal sequence of the seismic audio, and s b [q] represents the qth signal value in the bth frame of the signal sequence s of the earthquake audio. Preferably, the earthquake events misjudged by the earthquake observation point using the outlier detection method include:

[0022] The following formula is used to clean the seismic monitoring data collected at the seismic observation points and to correct the earthquake events that are misjudged at the seismic observation points:

[0023]

[0024] Where D is a set of earthquake observation points, object d is the earthquake observation point to be detected as an outlier, object d′ is other earthquake observation points in D, r is the neighborhood, dist(d, d′) represents the distance from d to d′, and a (0≤a≤1) is the score threshold.

[0025] Prioritize the use of inverse distance weighted interpolation to obtain the status of the deployment points of urban surveillance cameras:

[0026]

[0027] Among them, Z(x) represents the estimated value of the observation point x where it is impossible to determine whether an earthquake has occurred, and Z i is the value of the i-th observation point where an earthquake has been determined, D i Represents the distance from the unknown point x to the i-th known point, P is the power parameter, and n is the total number of known points.

[0028] Prioritize the mel spectrum map and the zero-crossing rate map to obtain fused features, including:

[0029] F=S Mel-Spec +ZCR

[0030] Among them, F is the fusion feature, S Mel-Spec is the Mel spectrum of the seismic audio, and ZCR is the zero crossing rate map of the seismic audio.

[0031] Prioritize the use of spatial interpolation to analyze earthquake events identified at multiple seismic observation points to obtain information on the spatiotemporal distribution and evolution of earthquakes, including:

[0032]

[0033] Among them, t i represents the i-th moment, It represents the temporal and spatial distribution and evolution information of earthquakes from time t1 to time t2. represents the earthquake status at time t2, Indicates the earthquake status at time t1.

[0034] Preferably, the present invention provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of any one of the methods described in the first aspect when executing the program.

[0035] Preferably, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any one of the methods described in the first aspect.

[0036] The beneficial effects achieved by the present invention are:

[0037] The present invention firstly models the earthquake sound in the monitoring audio and constructs the audio time-frequency domain features of the earthquake sound in the monitoring audio; it analyzes the monitoring audio by combining multi-seismic audio feature fusion with deep learning to achieve the purpose of identifying earthquakes; and it realizes the high-temporal-spatial resolution perception of earthquakes through the collaborative work of multiple urban monitoring cameras in the monitoring sensor network.

[0038] The method of the present invention is a new type of earthquake monitoring method. It is based on the urban monitoring sensor network, with earthquake observation points widely distributed; based on the quantification of audio signal characteristics, there are no special requirements for earthquake monitoring equipment; based on the all-weather observation of 5G communication technology, it improves the transmission and processing rate of large-scale real-time earthquake data; multiple urban monitoring cameras work together to achieve point-to-surface monitoring, which improves the accuracy of the results. Therefore, this method provides technical support for high-temporal and spatial resolution monitoring of urban earthquakes, and has significant application value and social benefits.

[0039] Urban surveillance cameras are widely distributed and can be used as earthquake observation points, which improves the spatial density of monitoring and the spatial resolution of earthquake data. At the same time, the urban monitoring sensor network is based on 5G communication technology to observe all-weather, transmit earthquake data at high speed, improve the temporal resolution, and enhance the city's earthquake monitoring and disaster warning capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solution of the present application, the drawings required for use in the embodiments are briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0041] Figure 1 is a frequency domain characteristic diagram of earthquake audio in some embodiments of the present application;

[0042] Figure 2 is a time domain characteristic diagram of earthquake audio in some embodiments of the present application;

[0043] Figure 3 is a flow chart of some embodiments of the present application;

[0044] Figure 4 is a structural diagram of a deep learning model in some embodiments of the present application;

[0045] Figure 5 It is a diagram of the earthquake monitoring system of urban monitoring big data in some embodiments of the present application. DETAILED DESCRIPTION

[0046] To facilitate the technical solution of the application, some concepts involved in the application are first explained below.

[0047] See also Figure 1The present invention provides an earthquake recognition deep learning model EDSA based on the urban surveillance camera sensor network. The model includes the following steps. Figure 3 :

[0048] (1) Seismic data acquisition and preprocessing;

[0049] (2) Extraction and fusion of acoustic and audio features for earthquake monitoring;

[0050] (3) Build a deep learning model;

[0051] (4) Collaborative data analysis of surveillance cameras in multiple cities.

[0052] Urban surveillance cameras are widely distributed and can be used as earthquake observation points, which improves the spatial density of monitoring and the spatial resolution of earthquake data. At the same time, the urban monitoring sensor network is based on 5G communication technology to observe all-weather, transmit earthquake data at high speed, improve the temporal resolution, and enhance the city's earthquake monitoring and disaster warning capabilities.

[0053] The present invention first models the earthquake sound in the monitoring audio and constructs the audio time-frequency domain features of the earthquake sound in the monitoring audio. Figure 1 and Figure 2 ; Use a combination of multi-seismic audio feature fusion and deep learning to analyze monitoring audio to achieve the purpose of identifying earthquakes; through the collaborative work of multiple monitoring city monitoring cameras in the monitoring sensor network, high-temporal and spatial resolution perception of earthquakes can be achieved.

[0054] (1) Seismic data acquisition and preprocessing

[0055] The step (1) of seismic data acquisition and preprocessing specifically includes the following steps:

[0056] (1-1) Establish a connection with the urban monitoring sensor network through 5G communication technology to obtain a data set including earthquake audio, negative sample audio similar to the earthquake audio, and monitoring location information.

[0057] (1-2) Cut the earthquake audio and negative sample audio into 10s segments respectively

[0058] Natural sounds such as cat calls and car horns, the car horns sound for a short time of about 2 to 3 seconds, while after an earthquake, the vibration observed on the surface lasts about 10 to 15 seconds. After a large number of model training tests, the present invention found that the 10-second segment is conducive to improving the accuracy of the model.

[0059] (1-3) Normalize and filter the audio signal

[0060] Since the parameters of the urban surveillance cameras in the urban surveillance camera sensor network are different and the sound fields they are in are different, normalization processing is required to normalize the earthquake audio and negative sample audio. In addition, the quality of the earthquake audio and negative sample audio has a great impact on the accuracy of the model, and filtering processing is required to reduce the noise impact of the earthquake audio and negative sample audio to improve the data quality. The earthquake audio and negative sample audio are normalized and filtered. The specific program code is as follows:

[0061] sample_rate,audio=sf.read(audio_path)

[0062] audio_normalized=audio / np.max(np.abs(audio))

[0063] low_cutoff=180

[0064] high_cutoff=2500

[0065] filter_order=3

[0066] b,a=librosa.filters.bandpass(low_cutoff,high_cutoff,sr=sample_rate,order=filter_order)audio_filtered=librosa.filters.filter_bandpass(audio_normalized,sample_rate,b,a)

[0067] (1-4) Create a data set

[0068] The processed earthquake audio and negative sample audio are saved in folders representing different categories to form a data set. At the same time, the audio category of the earthquake audio, the audio category of the negative sample audio, and the monitoring location information are written into the label file. The earthquake audio is marked as 1, and other categories of audio such as negative sample audio are marked as 0.

[0069] (2) Earthquake monitoring audio and video feature extraction and fusion

[0070] A large number of experiments have found that the Mel-spectrogram can well describe the frequency domain characteristics of seismic audio, and the zero-crossing rate graph can well describe the time domain characteristics of seismic audio.

[0071] The step (2) specifically comprises the following steps:

[0072] (2-1) Read the tag file created in (1-4) and load audio files in batches;

[0073] (2-2) Create a Mel-spectrogram of audio

[0074] The descriptive mathematical formula for the process of making a mel-spectrogram is:

[0075] S Mel-Spec (t,f)=LogScale(MelFilterBank(FFT(Window(s(t))))) (1)

[0076] Where S Mel-Spec (t,f) represents the Mel spectrogram function, where t is time and f is frequency. LogScale represents the logarithmic transformation of the output of the Mel filter. MelFilterBank represents the processing of the FFT result through the Mel filter bank, mapping the frequency domain to the Mel frequency scale. FFT represents the fast Fourier transform of the signal frame after the window function.

[0077] Window(s(t)) means dividing the signal s(t) into frames and applying the window function to each frame. The core code is as follows:

[0078] n_mels=128

[0079] print(file_path)

[0080] mel=librosa.feature.melspectrogram(S=stft,n_mels=n_mels)

[0081] mel_db=librosa.amplitude_to_db(abs(mel))

[0082] normalized_mel=librosa.util.normalize(mel_db)

[0083] plt.figure(figsize=(12,4))

[0084] librosa.display.specshow(mel_db,x_axis='time',y_axis='mel')

[0085] plt.colorbar(format='%+2.0f dB');

[0086] plt.title('MEL-Scaled Spectrogram')

[0087] plt.tight_layout()

[0088] plt.show()

[0089] (2-3) Create a zero-crossing rate graph of earthquake audio

[0090] The descriptive mathematical formula for the process of making a zero-crossing rate graph is:

[0091]

[0092] Where ZCR(b) represents the zero-crossing rate graph of the b-th frame, N represents the total number of signal values ​​in the b-th frame of the seismic audio, sign(s) represents the sign function, s represents the signal sequence of the seismic audio, and s b [q] represents the qth signal value in the bth frame of the signal sequence s of the seismic audio.

[0093]

[0094] The core code is:

[0095] zero_crossings=np.where(np.diff(np.sign(audio_signal)))[0]

[0096] zcr=len(zero_crossings) / len(audio_signal)

[0097] plt.figure(figsize=(10,6))

[0098] time=np.arange(len(audio_signal)) / sample_rate

[0099] plt.plot(time,audio_signal,color="blue")

[0100] plt.plot(time[zero_crossings],audio_signal[zero_crossings],'ro',color="blue")

[0101] plt.title('Audio Signal with Zero Crossings')

[0102] plt.xlabel('Time(s)')

[0103] plt.ylabel('Amplitude')

[0104] plt.savefig("1=zcr.png",dpi=300)

[0105] plt.show()

[0106] (2-4) Audio feature fusion

[0107] Feature fusion formula:

[0108] F=S Mel-Spec +ZCR (4)

[0109] Where F is the fused feature, S Mel-Spec is the Mel spectrum of the seismic audio, and ZCR is the zero crossing rate map of the seismic audio.

[0110] (3) Build a deep learning model, see the attached Figure 4 , identify earthquake audio.

[0111] The step (3) specifically comprises the following steps:

[0112] (3-1) Build a Transformer model and input the fused features obtained in (2-4) into the Transformer model.

[0113] The Encoder layer of the Transformer model can extract image features (see Appendix for details). Figure 5 ).

[0114] The core code of this step is as follows:

[0115] self.patch_emb=nn.Conv2D(in_channels=3, out_channels=hidden_size, kernel_size=patch_size, stride=patch_size)

[0116] self.pos_embed=self.create_parameter(shape=[1,input_size / / patch_siz

[0117] e,input_size / / patch_size,hidden_size],default_initializer=nn.initializer.Normal(std=0.05))

[0118] self.dropout=nn.Dropout(dropout)

[0119] layer=nn.TransformerEncoderLayer(hidden_size,num_heads,hidden_size*8,dropout)

[0120] self.transformer_encoder=nn.TransformerEncoder(layer,num_layers)

[0121] (3-2) Extract the feature tensor output by the Transformer model

[0122] self.label = []

[0123] self.tensors = []

[0124] B = img.shape(tensor)[0]

[0125] tensor=self.patch_emb(tensor)

[0126] cls=self.cls_token.expand((B,-1,-1))

[0127] tensor = tensor + self.pos_emb

[0128] tensor=self.pos_drop(tensor)

[0129] for b in self.blocks:

[0130] tensor = b(tensor)

[0131] self.tensors.add(tensor)

[0132] self.label.append(cls)

[0133] (3-3) Build an LSTM model, add a fully connected layer, and use the feature tensor extracted in (3-2) as input to further mine the features of the earthquake audio time series and realize the recognition of earthquake audio.

[0134] model=ImageLSTMClassifier(inputdim,hiddendim,numlayers,outputdim)

[0135] for epoch in range(numepochs):

[0136] for batchfeatures,batchlabels in dataloader:

[0137] op.zerograd()

[0138] outputs=model(batchfeatures)

[0139] loss=cr(outputs.squeeze(),batchlabels.float())

[0140] loss.backward()

[0141] op.step()

[0142] (4) Collaborative data analysis of surveillance cameras in multiple cities

[0143] (4-1) Cleaning earthquake monitoring data from earthquake observation points to ensure data quality

[0144] In actual applications, there will be individual outliers that misjudge earthquake events. In order to study the temporal and spatial distribution of earthquakes, process evolution and other characteristics, it is necessary to correct the misjudged observation points to ensure data quality. Earthquake monitoring data refers to whether an earthquake occurs at a seismic observation point. The outliers are corrected using a distance-based outlier detection method. For a set of data objects D, a distance threshold r is specified to define the neighborhood of the object. Suppose object d, d′∈D, consider the number of other data objects in the r domain of d. If most objects in D are far away from d, then d is considered an outlier. Let r (r≥0) be the distance threshold, a (0≤a≤1) be the score threshold, dist(d,d′) represents the distance from d to d′, and if d satisfies Equation 6, then d is an outlier.

[0145]

[0146] (4-2) The spatial interpolation method is used to analyze earthquake events identified at multiple earthquake observation points, and to achieve earthquake spatiotemporal distribution images and process evolution images as well as quantitative analysis.

[0147] Through the inverse distance weighted interpolation method, the unknown point status can be obtained and the data features can be enriched. The formula is as follows:

[0148]

[0149] Where Z(x) represents the estimated value of the unknown point x. i is the value of the i-th known point. iRepresents the distance from the unknown point x to the i-th known point. P is the power parameter that controls the decay rate of the distance weight. n is the total number of known points.

[0150] By analyzing the judgment of earthquake events by urban surveillance cameras at different time points, the temporal and spatial distribution and evolution process of earthquakes can be obtained, and then quantitatively analyzed, the following formula is obtained:

[0151] ΔMap m-z =Map m -Map z , (7),

[0152] Where m and z represent two moments in the earthquake process, and moment m is later than moment z. m-z Indicates the temporal and spatial distribution and evolution information of earthquakes from time z to time m. m ,Map z Represent the earthquake status at time m and time z respectively.

[0153] In an embodiment of the present application, the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of any of the above methods when executing the program.

[0154] In an embodiment of the present application, the present invention provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of any of the above methods are implemented.

[0155] The method of the present invention is a new type of earthquake monitoring method. It is based on the urban monitoring sensor network, with earthquake observation points widely distributed; based on the quantification of audio signal characteristics, there are no special requirements for earthquake monitoring equipment; based on the all-weather observation of 5G communication technology, it improves the transmission and processing rate of large-scale real-time earthquake data; multiple urban monitoring cameras work together to achieve point-to-surface monitoring, which improves the accuracy of the results. Therefore, this method provides technical support for high-temporal and spatial resolution monitoring of urban earthquakes, and has significant application value and social benefits.

[0156] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0157] Those skilled in the art will readily appreciate other embodiments of the present invention after considering the specification and practicing the invention invented herein. This application is intended to cover any variations, uses or adaptations of the present invention that follow the general principles of the present invention and include common knowledge or customary techniques in the art that are not invented by the present invention. The specification and examples are to be considered exemplary only, and the true scope and spirit of the present invention are indicated by the following claims.

[0158] The above specific implementation methods further illustrate the purpose, technical solutions and beneficial effects of the present application in detail. It should be understood that the above are only specific implementation methods of the present application and are not used to limit the protection scope of the present application. Any modifications, equivalent substitutions, improvements, etc. made on the basis of the technical solutions of the present application should be included in the protection scope of the present application.

Claims

1. A method for earthquake identification based on a city surveillance camera sensor network, characterized in that: include: Obtain earthquake monitoring data collected from multiple earthquake observation points, including earthquake audio, negative sample audio, and monitoring location information; Normalize and filter the earthquake audio and negative sample audio; Write the audio category of the earthquake audio, the audio category of the negative sample audio, and the monitoring location information into the label file; Extract the frequency domain features of the earthquake audio and obtain the Mel spectrum graph; Extract the time domain features of the seismic audio and obtain the zero-crossing rate map; Fuse the Mel-spectrogram and the zero-crossing rate map to obtain fusion features; Input the obtained fusion features into the Transformer model to obtain the feature tensor; Input the feature tensor into the improved LSTM model to obtain the features of the earthquake audio time series and realize the recognition of earthquake audio; The outlier detection method is used to correct the earthquake events that are misjudged at the earthquake observation point, and a single surveillance camera is used as a earthquake observation point; The spatial interpolation method is used to analyze earthquake events identified at multiple earthquake observation points to obtain information on the spatiotemporal distribution and evolution of earthquakes.

2. The earthquake identification method based on the urban monitoring camera sensor network according to claim 1 is characterized in that: The calculation formula of the Mel spectrum is: S Mel-Spec (t,f)=LogScale(MelFilterBank(FFT(Window(s(t))))), Among them, S Mel-Spec (t,f) represents the Mel-spectrogram function, t is time, f is frequency, LogScale represents the logarithmic transform of the output of the Mel filter, MelFilterBank represents the result of FFT processing through the Mel filter bank, FFT represents the fast Fourier transform of the signal frame after the window function, and Window(s(t)) represents framing the signal s(t) and applying the window function to each frame.

3. The earthquake identification method based on the urban monitoring camera sensor network according to claim 1 is characterized in that: The calculation formula of the zero-crossing rate map is: Where ZCR(b) represents the zero-crossing rate graph of the b-th frame, N represents the total number of signal values ​​in the b-th frame of the seismic audio, sign(s) represents the sign function, s represents the signal sequence of the seismic audio, and s b [q] represents the qth signal value in the bth frame of the signal sequence s of the seismic audio.

4. The earthquake identification method based on the urban monitoring camera sensor network according to claim 1 is characterized in that: Earthquake events that are misjudged by earthquake observation points using the outlier detection method include: The following formula is used to clean the seismic monitoring data collected at the seismic observation points and to correct the earthquake events that are misjudged at the seismic observation points: Where D is a set of earthquake observation points, object d is the earthquake observation point to be detected as an outlier, object d′ is other earthquake observation points in D, r is the neighborhood, dist(d, d′) represents the distance from d to d′, a (0≤a≤1) is the score threshold, and earthquake monitoring data refers to whether an earthquake occurs.

5. The earthquake identification method based on the urban monitoring camera sensor network according to claim 4 is characterized in that: Using the inverse distance weighted interpolation method, the status of the deployment points of urban surveillance cameras is obtained: Among them, Z(x) represents the estimated value of the observation point x where it is impossible to determine whether an earthquake has occurred, and Z i is the value of the i-th observation point where an earthquake has been determined, D i Represents the distance from the unknown point x to the i-th known point, P is the power parameter, and n is the total number of known points.

6. The earthquake identification method based on the urban monitoring camera sensor network according to claim 1 is characterized in that: The Mel spectrum graph and the zero-crossing rate graph are fused to obtain fusion features, including: F=S Mel-Spec +ZCR, Among them, F is the fusion feature, S Mel-Spec is the Mel spectrum of the seismic audio, and ZCR is the zero crossing rate map of the seismic audio.

7. The earthquake identification method based on the urban monitoring camera sensor network according to claim 1 is characterized in that: The spatial interpolation method is used to analyze earthquake events identified at multiple earthquake observation points to obtain information on the temporal and spatial distribution and evolution of earthquakes, including: ΔMap m-z =Map m -Map z , Among them, m and z represent two moments in the earthquake process, and moment m is later than moment z. ΔMap m-z Indicates the temporal and spatial distribution and evolution information of earthquakes from time z to time m. m ,Map z Represent the earthquake status at time m and time z respectively.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the method according to any one of claims 1 to 7 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images