Comprehensive pipe gallery risk identification method and system based on sound signal imaging

Abstract

The present application proposes a method and system for identifying risks in an integrated pipe corridor based on the visualization of sound signals, wherein the method includes: collecting multiple types of disaster sound signals inside the integrated pipe corridor as sample sound signals, wherein the sample sound signals are stored in the sound information disaster library of the integrated pipe corridor; processing the sample sound signals into images to generate a sound spectrogram, and using it as a sample image, annotating the sample image with a detection frame based on the disaster information corresponding to the sample sound signal; training a disaster identification model based on a deep learning target detection algorithm based on the sample image and the corresponding annotated detection frame to generate a trained target disaster identification model; collecting the current sound signal sequence to be identified in the integrated pipe corridor and generating a sound spectrogram to be identified, and inputting the sound spectrogram to be identified into the target disaster identification model to output the corresponding disaster identification result. The integrated pipe corridor risk identification method is improved, and the timeliness and accuracy of risk identification are improved.

Description

Technical Field

[0001] The present application relates to the field of sound signal processing, and in particular to a method and system for identifying risks in an integrated pipe gallery based on sound signal visualization. Background Art

[0002] With the development of cities, underground pipelines play an important role in the operation and maintenance of modern cities. With the increase in the number of underground pipelines, the urban pipeline corridor method is adopted in related technologies to achieve the management of underground pipelines. Therefore, risk identification and early warning of urban pipeline corridors are becoming increasingly important.

[0003] In the related technology, methods such as video monitoring and sensor sensing are used, but the monitoring methods are not perfect. Summary of the invention

[0004] This application aims to solve one of the technical problems in the related art to a certain extent. To this end, one purpose of this application is to propose a comprehensive pipeline corridor risk identification method and system based on sound signal visualization, so as to solve the problem of imperfect risk identification and early warning methods for urban pipeline corridors to a certain extent. The technical solution of this application is as follows:

[0005] In a first aspect, the present application proposes a method for identifying risks in an integrated pipe corridor based on the visualization of sound signals, comprising the following steps: collecting sound signals corresponding to multiple types of disasters inside the integrated pipe corridor as sample sound signals, wherein the sample sound signals are stored in a sound information disaster library of the integrated pipe corridor; visualizing the sample sound signals to generate a sound spectrogram of the sample sound signals, and using the sound spectrogram as a sample image of the sample sound signal, and annotating the sample image with a detection frame based on the disaster information corresponding to the sample sound signal; training a disaster identification model constructed based on a deep learning target detection algorithm based on the sample image and the corresponding annotated detection frame to generate a trained target disaster identification model; collecting a sound sequence to be identified in the integrated pipe corridor, wherein the sound sequence may include multiple sound signals to be identified; generating a sound spectrogram to be identified of the sound sequence to be identified, and inputting the sound spectrogram to be identified into the target disaster identification model to output a disaster identification result corresponding to the sound sequence to be identified.

[0006] In addition, the integrated pipe gallery risk identification method based on sound signal visualization proposed in the first aspect of the present application also has the following additional features:

[0007] According to one embodiment of the present application, the sound signals corresponding to various types of disasters inside the integrated pipeline corridor are collected as sample sound signals, including: collecting real sound signals before and during the occurrence of various types of disasters in the integrated pipeline corridor; simulating the sounds before and during the occurrence of various types of disasters in the integrated pipeline corridor to generate simulated sound signals; and generating the sample sound signals based on the real sound signals and the simulated sound signals.

[0008] According to one embodiment of the present application, generating the sample sound signal based on the real sound signal and the analog sound signal includes: performing pre-emphasis processing operations, frame processing operations, windowing processing operations and noise reduction processing operations on the real sound signal and the analog sound signal to generate the sample sound signal.

[0009] According to one embodiment of the present application, the method further includes: generating a corresponding virtual sound spectrogram based on the sound spectrogram of the sample sound signal; using the sound spectrogram and the virtual sound spectrogram as the sample image, and annotating the sample image with a detection box according to the disaster information corresponding to the sample sound signal.

[0010] According to one embodiment of the present application, the network structures such as the feature extraction skeleton, detection layer, and fusion layer of the disaster recognition model based on the deep learning target detection algorithm are determined according to the feature complexity of the sample image and the size distribution of the detection box annotation results.

[0011] According to one embodiment of the present application, the generating of the sound spectrogram to be identified of the sound sequence to be identified, and inputting the sound spectrogram to be identified into the target disaster identification model to output the disaster identification result corresponding to the sound sequence to be identified, includes: inputting the sound spectrogram to be identified into the target disaster identification model to output the predicted detection box of the sound spectrogram to be identified; obtaining the labeled detection box i missed by the target disaster identification model, and obtaining the predicted detection box i-1 and predicted detection box i+1 adjacent to the labeled detection box i; based on the attribute information of the predicted detection box i-1 and the predicted detection box i+1, determining the attribute information of the predicted detection box i corresponding to the labeled detection box i; according to the attribute information of the predicted detection box i, outputting the disaster identification result corresponding to the sound sequence to be identified that was missed by the target disaster identification model.

[0012] According to one embodiment of the present application, the method also includes: after disaster identification, obtaining attribute information of the target detection box on the sound spectrogram to be identified; and determining the average length, average height and average interval length of the target detection box based on the attribute information of each target detection box.

[0013] According to one embodiment of the present application, the method also includes: after disaster identification, obtaining the size of the sound spectrogram to be identified; predicting the average duration of the target sound signal when the disaster occurs based on the average length and the size; obtaining the frequency band range where the target sound signal is located based on the attribute information of the target detection frame and the size; and obtaining the average interval duration of adjacent target sound signals based on the average interval length and the size.

[0014] The second aspect of the present application proposes a comprehensive pipe corridor risk identification system based on sound signal visualization, comprising: a sound signal acquisition module, used to collect sound signals corresponding to multiple types of disasters inside the comprehensive pipe corridor as sample sound signals, wherein the sample sound signals are stored in a comprehensive pipe corridor disaster sound library; an image conversion module, used to image the sample sound signal to generate a sound spectrogram of the sample sound signal, and use the sound spectrogram as a sample image of the sample sound signal, and annotate the sample image with a detection frame based on the disaster information corresponding to the sample sound signal; a model training module, used to train a disaster identification model constructed based on a deep learning target detection algorithm based on the sample image and the corresponding annotated detection frame to generate a trained target disaster identification model; the sound signal acquisition module, also used to collect a sound signal sequence to be identified in the comprehensive pipe corridor, wherein the sound signal sequence includes multiple sound signals to be identified; a risk assessment module, used to generate a sound spectrogram to be identified of the sound signal to be identified, and input the sound spectrogram to be identified into the target disaster identification model to output a disaster identification result corresponding to the sound signal to be identified.

[0015] The present application proposes a method and system for identifying risks in an integrated pipe corridor based on the visualization of sound signals. The sound signals corresponding to various types of disasters that may exist in the integrated pipe corridor collected are used as sample generation signals, and a sound spectrogram corresponding to the sample sound signal is generated. The sound spectrogram is used as a sample image corresponding to the sample sound signal, and the sound spectrogram paragraphs corresponding to the sound signals of different types of disasters in the sample image are marked with detection frames. According to the sample image and the detection frame marked therein, the constructed disaster identification model to be trained is trained to generate a target disaster identification model. Further, a sound spectrogram of each sound signal to be identified in the sound signal sequence to be identified collected by the integrated pipe corridor is generated, and it is input into the trained target disaster identification model for disaster identification, and then the disaster identification result corresponding to the sound signal to be identified is output. In the present application, by collecting and analyzing the sound signals in the integrated pipe corridor, the identification and early warning of possible disasters in the integrated pipe corridor are realized, the method of risk identification of the integrated pipe corridor is improved, and the timeliness and accuracy of risk identification of the integrated pipe corridor are effectively improved.

[0016] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings herein are incorporated into the specification and constitute a part of the specification, illustrate embodiments consistent with the present application, and together with the specification are used to explain the principles of the present application, and do not constitute improper limitations on the present application.

[0018] Figure 1 A schematic diagram of a flow chart of a method for identifying risks of an integrated pipe gallery based on sound signal visualization according to an embodiment of the present application;

[0019] Figure 2 A schematic diagram of a flow chart of a method for identifying risks of an integrated pipe gallery based on sound signal visualization according to another embodiment of the present application;

[0020] Figure 3 A schematic diagram of a flow chart of a method for identifying risks of an integrated pipe gallery based on sound signal visualization according to another embodiment of the present application;

[0021] Figure 4 A schematic diagram of a flow chart of a method for identifying risks of an integrated pipe gallery based on sound signal visualization according to another embodiment of the present application;

[0022] Figure 5 A schematic diagram of a flow chart of a method for identifying risks of an integrated pipe gallery based on sound signal visualization according to another embodiment of the present application;

[0023] Figure 6 A schematic diagram of a flow chart of a method for identifying risks of an integrated pipe gallery based on sound signal visualization according to another embodiment of the present application;

[0024] Figure 7 This is a structural schematic diagram of a comprehensive pipe gallery risk identification system based on sound signal visualization according to an embodiment of the present application;

[0025] Figure 8 This is a structural schematic diagram of a comprehensive pipe gallery risk identification system based on sound signal visualization according to another embodiment of the present application;

[0026] Fig. 9 A block diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0027] In order to enable ordinary persons in the art to better understand the technical solutions of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings. The same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0028] Figure 1 FIG. 1 is a flow chart of a method for identifying risks of an integrated pipe gallery based on sound signal visualization according to an embodiment of the present application. Figure 1 As shown, the method includes:

[0029] S101, collecting sound signals corresponding to various types of disasters inside the utility corridor as sample sound signals, wherein the sample sound signals are stored in a utility corridor sound information disaster database.

[0030] In practice, the underground pipelines in the comprehensive pipeline corridor can undertake multiple functions such as urban information transmission, energy transportation, flood drainage and disaster reduction, waste disposal, etc. Therefore, in the daily operation of underground pipelines, pipelines with different functions are likely to suffer different types of disasters.

[0031] For example, underground pipelines that are responsible for energy transmission may have different disasters depending on the energy they transmit. Cables responsible for transmitting electric energy resources may have cable discharge disasters, and pipelines responsible for transmitting gas energy may have gas leakage disasters, etc.

[0032] Furthermore, different types of disasters produce different sound signals during their occurrence. Therefore, the identification and early warning of possible disasters in the integrated utility tunnel can be achieved based on the characteristics of different sound signals.

[0033] Optionally, based on the sound collection equipment arranged in the integrated pipe corridor, sound signals corresponding to various types of disasters occurring in the integrated pipe corridor can be collected and determined as sample sound signals.

[0034] Furthermore, a comprehensive pipeline corridor sound information disaster database can be constructed, and the collected sound signals corresponding to various disasters that may occur inside the comprehensive pipeline corridor can be stored in the constructed comprehensive pipeline corridor sound information disaster database to facilitate subsequent calling and updating of sample sound signals.

[0035] S102, performing image processing on the sample sound signal to generate a sound spectrogram of the sample sound signal, and using the sound spectrogram as a sample image of the sample sound signal, and annotating the sample image with a detection frame based on the disaster information corresponding to the sample sound signal.

[0036] In the implementation, due to the propagation characteristics of the sound signal, it is difficult to obtain the characteristics of the sound signal. Therefore, the sound signal can be processed into an image and then converted into an image that can present the characteristics of the sound signal, so that the characteristics of the sound signal can be more accurately extracted and identified based on the image.

[0037] Optionally, the sample sound signal can be converted into a corresponding sound spectrogram, wherein the sound spectrogram can present the form and characteristics corresponding to the sample sound signal. The sound signals corresponding to different types of disasters have different forms and characteristics presented in the sound spectrogram. Therefore, the features of the sound signals corresponding to different types of disasters can be extracted according to the forms and characteristics of the sound signals in the sound spectrogram.

[0038] It should be noted that, in addition to the sound spectrogram, the method of visualizing the sound signal may include other processing methods that can present the characteristics of the sound signal, which is not limited here.

[0039] Among them, the horizontal axis of the sound spectrogram represents time, the vertical axis represents frequency, and the color of the coordinate point represents energy.

[0040] Furthermore, the sound spectrogram can be used as a sample image corresponding to the sample sound signal, and based on the analysis of the expression forms and characteristics of different types of sound signals in the sound spectrogram, the forms and characteristics of different types of sound signals in the sample image can be obtained.

[0041] Based on the manifestation forms and characteristics of different types of sound signals in the sound spectrogram, the target forms and target characteristics of sample sound signals corresponding to different types of disasters in the sample images are determined and annotated.

[0042] Among them, the target forms corresponding to the sample sound signals of different types of disasters can be marked in the sample image through a rectangular detection frame, so as to generate a sample image with the detection frame marked.

[0043] Optionally, the sample image can be annotated with a detection frame through a set software, for example, through a visual image calibration tool (LabelImg software), the target shape corresponding to the fire in the sample image is marked with the label fire, the target shape corresponding to the gas leakage is marked with the label leakage, and the target shape corresponding to the cable discharge is marked with the label discharge.

[0044] S103, based on the sample image and the corresponding annotated detection frame, the disaster recognition model constructed based on the deep learning target detection algorithm is trained to generate a trained target disaster recognition model.

[0045] In implementation, the disaster identification model can be used to identify and warn of different types of disasters that may occur in the integrated pipeline corridor.

[0046] Furthermore, in order to make the disaster identification model more efficient and accurate in actual application, it is necessary to train the disaster identification model based on the deep learning target detection algorithm so that the performance of the disaster identification model based on the deep learning target detection algorithm can be optimized.

[0047] In an embodiment of the present application, the constructed disaster identification model can be trained through labeled sample images. Through multiple rounds of training, the disaster identification model based on the deep learning target detection algorithm can realize the identification of different types of disasters existing in the integrated pipeline corridor, and then generate a trained target disaster identification model.

[0048] Optionally, the loss function of the disaster identification model can be obtained based on the relevant information of the predicted detection box in the output results of the disaster identification model and the relevant information of the labeled detection box in the sample image after each round of model training. Based on the loss function, the parameters of the constructed disaster identification model can be adjusted to optimize the disaster identification model.

[0049] Furthermore, after each round of training of the disaster identification model is completed, it can be determined whether the disaster identification model of the current round meets the end condition of the model training. If the end condition of the model training is met, the model training is terminated and the target disaster identification model is generated. If the end condition of the model training is not met, the weight parameters of the disaster identification model are adjusted based on the loss results of the current round of training until the end condition of the training is met.

[0050] Among them, the corresponding model training end condition can be set based on the model accuracy of the disaster identification model, and the corresponding model training end condition can also be set based on the training rounds of the disaster identification model.

[0051] It should be noted that during the training of the disaster identification model, the model parameters of the disaster identification model can be adjusted as a whole after the set round of iterations. For example, the input resolution and batch size can be changed every 10 rounds to improve the training efficiency.

[0052] Furthermore, an optimal weight configuration of the target disaster identification model is generated.

[0053] S104, collecting the sound sequence to be identified in the integrated pipe gallery, where the sound sequence includes multiple sound signals to be identified.

[0054] Furthermore, the trained target disaster recognition model can be applied in actual scenarios, and the sound signals of the integrated pipeline corridor in the actual scenario can be collected to generate a sound sequence to be identified.

[0055] Among them, the collected sound sequence includes multiple sound signals to be identified in the integrated pipeline corridor.

[0056] S105, generating a sound spectrogram to be identified corresponding to the sound sequence to be identified, and inputting the sound spectrogram to be identified into a target disaster identification model to output a disaster identification result corresponding to the sound sequence to be identified.

[0057] In an embodiment of the present application, in order to realize the recognition of multiple disaster sound signals in the current sound sequence to be recognized based on the trained target disaster recognition model, optionally, a sound spectrogram corresponding to the sequence can be generated based on the current sound sequence to be recognized and determined as the sound spectrogram to be recognized.

[0058] Furthermore, the sound spectrogram to be identified is input into the trained target disaster identification model, the target forms of different disaster types in the sound spectrogram to be identified are identified by the model, and the detection result image is output to generate the target disaster identification result.

[0059] Among them, the risk levels of different types of disasters can also be divided based on the target disaster model identification results, so that the output results of the target disaster identification model are more intuitive. Among them, the disaster risk level can be divided into numerical levels, and the higher the corresponding numerical level, the greater the risk.

[0060] The present application proposes a method for identifying risks in an integrated pipe corridor based on the visualization of sound signals. The sound signals corresponding to various types of disasters that may exist in the integrated pipe corridor collected are used as sample sound signals, and a sound spectrogram corresponding to the sample sound signals is generated. The sound spectrogram is used as a sample image corresponding to the sample sound signal, and the target morphology corresponding to the sound signals of different types of disasters in the sample image is marked with a detection frame. According to the sample image and the detection frame marked therein, the disaster identification model constructed based on the deep learning target detection algorithm is trained to generate a target disaster identification model. Further, a sound spectrogram of the sound signal sequence to be identified collected by the integrated pipe corridor is generated, and it is input into the trained target disaster identification model for disaster identification, and then the corresponding disaster identification result is output. In the present application, by collecting and analyzing the sound signals in the integrated pipe corridor, the identification and early warning of possible disasters in the integrated pipe corridor are realized, the method for risk identification of the integrated pipe corridor is improved, and the timeliness and accuracy of risk identification of the integrated pipe corridor are effectively improved.

[0061] In the above embodiment, the acquisition of the sample sound signal can be combined with Figure 2 Further understanding, Figure 2 This is a flow chart of a method for identifying risks of integrated pipe corridors based on sound signal visualization according to another embodiment of the present application. Figure 2 As shown, the method includes:

[0062] S201 collects real sound signals before and during the occurrence of various disasters in the integrated pipeline corridor.

[0063] In the actual integrated pipeline corridor scenario, different types of disasters are accompanied by different sounds. In order to accurately identify the risks in the integrated pipeline corridor, the real sound signals of various types of disasters in the integrated pipeline corridor during their occurrence can be collected.

[0064] The real sound signals collected during the occurrence of various types of disasters may include real sound signals before and during the occurrence of the disaster.

[0065] In implementation, the occurrence process of different types of disasters is accompanied by changes in sound. Therefore, based on the characteristics of the real sound signals collected at different occurrence stages, the change process of the real sound signals of different types of disasters during their occurrence can be obtained, thereby achieving accurate extraction of the sound signal characteristics of different types of disasters.

[0066] S202, simulating the sounds before and during the occurrence of various disasters in the integrated pipeline corridor to generate simulated sound signals.

[0067] In the implementation, since the frequency of occurrence of each type of disaster is relatively small in the actual scenario of the integrated pipeline corridor, the sample set generated based on the real sound signals corresponding to different types of disasters collected cannot meet the training requirements of the disaster recognition model based on the deep learning target detection algorithm.

[0068] In an embodiment of the present application, a corresponding simulation environment can be built based on the actual scenario of the integrated pipeline corridor, and different types of disasters that may occur in the integrated pipeline corridor can be simulated in the built simulation environment. During the occurrence of the simulated disaster, the accompanying sound signals can be collected to generate simulated sound signals corresponding to different types of disasters.

[0069] S203: Generate a sample sound signal based on the real sound signal and the simulated sound signal.

[0070] In an embodiment of the present application, the collected real sound signals and simulated sound signals can be further processed to achieve the purpose of training a disaster recognition model based on a deep learning target detection algorithm based on the collected sound signals.

[0071] Furthermore, pre-emphasis processing operations, frame processing operations, windowing processing operations and noise reduction processing operations may be performed on real sound signals and simulated sound signals.

[0072] Among them, the collected real sound signal and analog sound signal can be pre-emphasized, so as to emphasize the high-frequency part of the real sound signal and the analog sound signal, thereby increasing the high-frequency resolution of the real sound signal and the analog sound signal.

[0073] Optionally, the pre-emphasis processing operation on the real sound signal and the analog sound signal can be implemented by a first-order high-pass digital filter.

[0074] Furthermore, the real sound signal and the analog sound signal after pre-emphasis can be windowed and framed, and the collected real sound signal and the analog sound signal can be framed based on a movable window of a set length, thereby generating a number of short-segment sound signals.

[0075] For example, the frame length is set to 200 frames, the frame shift is set to 80 frames, and the collected real sound signal and simulated sound signal are framed using a Hamming window of a set length, thereby generating a number of short sound signals of 200 frames in length.

[0076] Furthermore, noise reduction processing may be performed on a number of short-segment sound signals generated based on the collected real sound signals and simulated sound signals.

[0077] The start and end points of each short segment of the sound signal can be determined by using a sound signal endpoint detection method. Optionally, the short-time energy and short-time zero-crossing rate methods can be used to detect the endpoints of the sound signal.

[0078] Furthermore, noise reduction processing is performed on the sound signal of each short segment with a determined start and end point.

[0079] Optionally, a wavelet threshold method may be used to implement noise reduction processing of the sound signal.

[0080] Furthermore, based on further processing operations on the collected real sound signal and the simulated sound signal, a corresponding sample sound signal is generated.

[0081] Optionally, a corresponding integrated pipe gallery disaster database based on sound information may be constructed, and the generated sample sound signals may be stored in a set location therein.

[0082] The integrated pipe corridor risk identification method based on sound signal visualization proposed in this application collects real sound signals before and during the occurrence of disasters from the actual scene of the integrated pipe corridor, and collects simulated sound signals before and during the occurrence of disasters from the simulated environment of the integrated pipe corridor, and generates corresponding sample sound signals by pre-processing the real sound signals and simulated sound signals such as pre-emphasis, windowing, framing, and noise reduction. In this application, the number of samples for disaster model training is effectively increased by collecting simulated sound signals, and the corresponding sample sound signals are generated by pre-processing the real sound signals and simulated sound signals, thereby effectively enhancing the accuracy of the results generated by subsequent operations based on the sample sound signals.

[0083] In the above embodiment, the sample image can also be combined with Figure 3 Further understanding, Figure 3 This is a flow chart of a method for identifying risks of integrated pipe corridors based on sound signal visualization according to another embodiment of the present application. Figure 3 As shown, the method includes:

[0084] S301, generating a corresponding virtual sound spectrogram based on the sound spectrogram of the sample sound signal.

[0085] In an embodiment of the present application, in order to extract the features of the sample sound signal more accurately and completely, it can be converted into a corresponding sound spectrogram, based on the extraction of the image morphology and features corresponding to the sound signals of different types of disasters in the sound spectrogram, thereby achieving the determination of the features of the sound signals of different types of disasters.

[0086] In the implementation, there may be a situation where the number of sound spectrograms generated based on the sample sound signal is small. Therefore, the sound spectrogram can be further processed to increase the number of sample images, thereby enabling the disaster recognition model based on the deep learning target detection algorithm to obtain effective model training.

[0087] Optionally, the sound spectrogram corresponding to the sample sound signal can be flipped, rotated and scaled, and the presentation angle of the sound spectrogram can be adjusted while retaining the target form and target features in the original sound spectrogram. Further, the flipped, rotated and scaled sound spectrogram can be determined as a virtual sound spectrogram.

[0088] Optionally, the part of the sound spectrogram corresponding to the sample sound signal that has the target form and characteristics can be intercepted and randomly spliced ​​with the sound spectrogram corresponding to the sound signal in the normal operation scenario of the integrated pipeline corridor, so as to generate the corresponding virtual sound spectrogram.

[0089] Furthermore, the sound spectrogram corresponding to the sample sound signal and the generated virtual sound spectrogram are determined as sample images for training a disaster recognition model based on a deep learning target detection algorithm.

[0090] S302, taking the sound spectrogram and the virtual sound spectrogram as sample images, and marking the sample images with detection frames according to the disaster information corresponding to the sample sound signals.

[0091] In an embodiment of the present application, in order to enable the sample images to realize effective training of the disaster recognition model, the target forms corresponding to the sound signals of different types of disasters in the sample images can be marked with detection boxes based on the disaster information corresponding to the sample sound signals.

[0092] It can be understood that, according to the target morphology and target features of the sound signals of different types of disasters in the sample sound signals shown in the sound spectrogram, the part of the sample image with similar morphology and features is determined and marked with a detection frame.

[0093] The integrated pipeline corridor risk identification method based on sound signal visualization proposed in this application effectively increases the number of sample images through further image processing of the sound spectrogram, thereby optimizing the training effect of the disaster identification model based on the deep learning target detection algorithm.

[0094] Furthermore, the output result of the target disaster identification model is corrected for missed detection, which can be combined with Figure 4 understand, Figure 4 This is a flow chart of a method for identifying risks of integrated pipe corridors based on sound signal visualization according to another embodiment of the present application. Figure 4 As shown, the method includes:

[0095] S401, inputting the sound spectrogram to be identified into the target disaster identification model to output a predicted detection box corresponding to the sound spectrogram to be identified.

[0096] It should be noted that the network structure of the constructed disaster identification model is similar to the target detection network (YOLOV3 network), target detection network (Region-CNN network), target detection network (Faster Region-CNN network), target detection network (SSD network), etc. First, it is a skeleton network composed of multiple feature extraction layers, and then passes through multiple convolutional layers, and then through multiple downsampling combinations, finally forming detection layers of different scales.

[0097] In the implementation, it is first necessary to statistically analyze the distribution of the annotation box sizes of the target forms corresponding to various types of disasters in the annotation results, and then determine the network structure of the disaster recognition model, including the feature extraction skeleton, detection layer, and fusion layer, based on the target size and feature complexity of the target in the sample image.

[0098] In the implementation, after the network structure of the disaster identification model is determined, the sample image and its corresponding annotation results are input into the disaster identification model to be trained, and then the disaster identification model to be trained is trained. After the training is completed, the target disaster identification model is obtained.

[0099] Furthermore, the sound spectrogram to be identified is input into the target disaster identification model, and the predicted detection box of the sound spectrogram to be identified is output.

[0100] During the acquisition of sound signals, the sound signals may be unstable, which may cause the target shape and features in the generated sound spectrogram to be identified to be fuzzy. Therefore, when the target disaster identification model identifies the sound spectrograms corresponding to the sound signals of different types of disasters, there is a possibility of missed detection.

[0101] Furthermore, the output results of the target disaster identification model can be corrected for missed detection, thereby achieving comprehensive and accurate identification of disasters.

[0102] Furthermore, since the target disaster recognition model realizes the target morphological recognition of different types of disasters in the sound spectrogram to be recognized by generating a prediction detection frame, the missed detection correction of the target disaster model can be realized based on the prediction detection frame output after the recognition is completed.

[0103] S402, obtaining a labeled detection box i that is missed by the target disaster recognition model, and obtaining a predicted detection box i-1 and a predicted detection box i+1 adjacent to the labeled detection box i.

[0104] In the embodiment of the present application, the missed detection in the spectrogram can be corrected based on the predicted detection box in the same sound spectrogram to be recognized.

[0105] In implementation, sound signals are continuously generated and propagated during the disaster process. Therefore, the target form and features of the sound signals of each type of disaster in the sound spectrogram to be identified have certain regularity and continuity.

[0106] It can be understood that the development of various disasters is relatively stable in a short period of time, and the target shapes of the corresponding sound signals presented in the sound spectrogram to be identified are similar in size and close in interval.

[0107] Furthermore, the missed detection situation can be corrected based on the above characteristics.

[0108] Optionally, by comparing the annotated detection box of the same sound spectrogram to be recognized with the predicted detection box output by the model, the missed annotated detection box i can be obtained, and the predicted detection boxes i-1 and i+1 adjacent to it can be obtained.

[0109] S403 , based on the attribute information of the predicted detection box i-1 and the predicted detection box i+1, determine the attribute information of the predicted detection box i corresponding to the labeled detection box i.

[0110] In the embodiment of the present application, after the predicted detection box i-1 and the predicted detection box i+1 are determined, the attribute information of the predicted detection box i corresponding to the missed labeled detection box i can be determined based on the attribute information of the two predicted detection boxes.

[0111] Optionally, since the target shape and features corresponding to the missed labeled detection frame i are similar to the target shapes and features in the predicted detection frame i-1 and the predicted detection frame i+1, the size information of the two predicted detection frames can be used to calculate the size information of the predicted detection frame i corresponding to the missed labeled detection frame i, and based on the coordinate information of the central pixel points of the two predicted detection frames, the coordinate information of the central pixel point of the predicted detection frame i corresponding to the missed labeled detection frame i can be calculated.

[0112] For example, assume that n labeled detection frames are annotated in the sample image, among which the target shape corresponding to the i-th labeled detection frame is missed.

[0113] Then the center pixel coordinates (x i ,y i ) is calculated as follows:

[0114]

[0115] Then the length L of the predicted detection box i corresponding to the i-th labeled detection box i The calculation formula is as follows:

[0116]

[0117] Then the height H of the predicted detection box i corresponding to the i-th labeled detection box i The calculation formula is as follows:

[0118]

[0119] S404: Outputting disaster recognition results in the sound sequence to be recognized that is missed by the target disaster recognition model according to the attribute information of the predicted detection box i.

[0120] In the embodiment of the present application, after determining the attribute information of the predicted detection box i corresponding to the marked detection box i, based on the attribute information of the predicted detection box i, the missed detection correction corresponding to the marked detection box i can be implemented.

[0121] Among them, the position and annotation range of the predicted detection box i are determined according to the coordinates of the central pixel point of the predicted detection box i, the length and height of the detection box, and then the missed detection correction of the target disaster recognition model is realized.

[0122] The integrated pipeline corridor risk identification method based on sound signal visualization proposed in this application inputs the sound spectrogram to be identified into the trained target disaster identification model, and corrects the possible missed detections of the target disaster identification model according to the output results, thereby effectively optimizing the detection effect of the algorithm and improving the recognition accuracy of the target disaster model.

[0123] Furthermore, the recognition results of the sound sequence to be recognized by the target disaster recognition model can be used to obtain the basic parameter indicators of the sound sequence to be recognized. Figure 5 understand, Figure 5 This is a flow chart of a method for identifying risks of integrated pipe corridors based on sound signal visualization according to another embodiment of the present application. Figure 5 As shown, the method includes:

[0124] S501, after disaster identification, obtain the attribute information of the target detection box on the sound spectrogram to be identified.

[0125] In an embodiment of the present application, disaster identification can be performed on sound signals in the integrated utility tunnel based on a trained target disaster identification model.

[0126] The current sound signal can be collected from the integrated pipe gallery to form a sound sequence to be identified, and then generate a corresponding sound spectrogram to be identified.

[0127] Furthermore, the sound spectrogram to be identified is input into the trained target disaster identification model, the disaster information in the sound spectrogram to be identified is identified based on the target disaster identification model, the predicted detection frame is output, and the recognition result is corrected to finally obtain the target detection frame.

[0128] After recognition, the attribute information of the target detection box on the sound spectrogram to be recognized can be obtained, including the coordinates, length, width, etc. of the central pixel point of each target detection box.

[0129] S502: Determine the average length, average height, and average interval length of each target detection frame according to the attribute information of each target detection frame.

[0130] In the embodiment of the present application, the sound signals corresponding to the target shapes and features in each target detection frame have their corresponding disaster types. Therefore, by calculating the attribute information of the target detection frames corresponding to the same disaster type, it is possible to predict the relevant information when each disaster type occurs.

[0131] Optionally, the duration and interval of each type of disaster may be predicted.

[0132] Furthermore, based on the length and height of each target detection frame, the interval length between the target detection frames, and the number of all target detection frames, the average length, average height, and average interval length of the target detection frames are obtained.

[0133] The size of the sound spectrogram to be recognized is set to S×S, its time range is 0~t, and its frequency range is 0~F. Based on the model recognition results and correction results, a total of n target detection boxes are marked in the sound spectrogram to be recognized.

[0134] The average length of the target detection box is average height and the average interval length The specific calculation formula is as follows:

[0135]

[0136] Wherein, n represents the frequency of occurrence of the target sound in the spectrogram of the sound to be recognized.

[0137] Furthermore, the size of the spectrogram of the sound to be recognized may be obtained.

[0138] Among them, based on the size of the sound spectrogram to be identified, the information of the target sound signal corresponding to various disasters can be obtained, including the time range and frequency range of the target sound signal.

[0139] Furthermore, the average duration of the target sound signal when a disaster occurs is predicted based on the average length and size.

[0140] Still based on the above settings, the size of the sound spectrogram to be recognized is S×S, its time range is 0~t, and its frequency range is 0~F. Based on the model recognition results and correction results, a total of n target detection boxes are marked in the sound spectrogram to be recognized.

[0141] When the disaster type corresponding to the target detection frame occurs, the average duration l of the target sound signal can be calculated by the following formula:

[0142]

[0143] Furthermore, the frequency band range of the target sound signal is obtained according to the attribute information and size of the target detection frame.

[0144] Still based on the above settings, the size of the sound spectrogram to be recognized is S×S, its time range is 0~t, and its frequency range is 0~F. Based on the model recognition results and correction results, a total of n target detection boxes are marked in the sound spectrogram to be recognized.

[0145] When the disaster type corresponding to the target detection frame occurs, the frequency band range of the target sound signal is f min ~f max , can be calculated by the following formula:

[0146]

[0147] Furthermore, the average interval duration of adjacent target sound signals is obtained according to the average interval length and size.

[0148] Still based on the above settings, the size of the sound spectrogram to be recognized is S×S, its time range is 0~t, and its frequency range is 0~F. Based on the model recognition results and correction results, a total of n target detection boxes are marked in the sound spectrogram to be recognized.

[0149] When the disaster type corresponding to the target detection frame occurs, the average interval duration t of the target sound signal can be calculated by the following formula:

[0150]

[0151] Furthermore, based on the disaster identification results and the above calculation results, risk level evaluation of different types of disasters can be achieved.

[0152] The proposed method for risk identification of utility corridors based on sound signal visualization inputs the sound spectrogram to be identified corresponding to the current sound sequence to be identified in the utility corridor into the trained target disaster identification model, and based on the target detection frame results obtained after identification and correction, the prediction of the relevant states when different types of disasters occur is realized. Based on the target disaster identification model, the efficiency and accuracy of disaster identification of utility corridors are effectively improved, thereby realizing effective risk identification of utility corridors.

[0153] Further, in order to better understand the above embodiments, Figure 6 , Figure 6 This is a flow chart of a method for identifying risks of integrated pipe corridors based on sound signal visualization according to another embodiment of the present application. Figure 6 As shown:

[0154] The real sound signals when disasters occur in the integrated pipe corridor and the simulated sound signals in the laboratory simulation environment are collected, and the collected real sound signals and simulated sound signals are pre-processed by pre-emphasis, windowing, framing and noise reduction to generate sample sound signals. The sample sound signal generates a corresponding sound spectrogram, and the sound spectrogram is determined as a sample image. The sample image is annotated with a detection frame according to the target morphology and features of the sound signals of different types of disasters in the sound spectrogram. The disaster recognition model constructed based on the deep learning target detection algorithm is trained based on the sample image and the annotated detection frame, and the model parameters are adjusted based on the results of each round of training output until the training is completed and a trained target disaster recognition model is generated. Furthermore, based on the trained target disaster recognition model, the sound sequence to be recognized in the integrated pipe corridor is identified for disasters, and the recognition results output by the target disaster recognition model are corrected for missed detection.

[0155] In this application, by collecting and analyzing the sound signals in the integrated pipeline corridor, the identification of possible disasters in the integrated pipeline corridor is achieved, the risk identification method of the integrated pipeline corridor is improved, and the timeliness and accuracy of the risk identification of the integrated pipeline corridor is effectively improved.

[0156] Corresponding to the integrated pipeline corridor risk identification methods based on sound signal visualization proposed in the above-mentioned embodiments, an embodiment of the present application also proposes an integrated pipeline corridor risk identification system based on sound signal visualization. Since the integrated pipeline corridor risk identification system based on sound signal visualization proposed in the embodiment of the present application corresponds to the integrated pipeline corridor risk identification methods based on sound signal visualization proposed in the above-mentioned embodiments, the implementation method of the above-mentioned integrated pipeline corridor risk identification method based on sound signal visualization is also applicable to the integrated pipeline corridor risk identification system based on sound signal visualization proposed in the embodiment of the present application, and will not be described in detail in the following embodiments.

[0157] Figure 7 The schematic diagram of the structure of the integrated pipe gallery risk identification system based on sound signal visualization according to an embodiment of the present application is as follows: the integrated pipe gallery risk identification system based on sound signal visualization 700 comprises a sound signal acquisition module 71, an image conversion module 72, a model training module 73, and a risk assessment module 74, wherein:

[0158] The sound signal collection module 71 is used to collect sound signals corresponding to various types of disasters inside the integrated pipe gallery as sample sound signals, wherein the sample sound signals are stored in the integrated pipe gallery sound information disaster database;

[0159] An image conversion module 72 is used to perform image processing on the sample sound signal to generate a sound spectrogram of the sample sound signal, and use the sound spectrogram as a sample image corresponding to the sample sound information, and perform detection frame annotation on the sample image based on the disaster information corresponding to the sample sound signal;

[0160] A model training module 73 is used to train the disaster recognition model based on the deep learning target detection algorithm based on the sample image and the corresponding annotated detection frame to generate a trained target disaster recognition model;

[0161] The sound signal collection module 71 is also used to collect the sound sequence to be identified in the integrated pipe gallery, and the sound sequence includes multiple sound signals to be identified;

[0162] The risk assessment module 74 is used to generate a sound spectrogram of the sound sequence to be identified, and input the sound spectrogram to be identified into the target disaster identification model to output a disaster identification result corresponding to the sound sequence to be identified.

[0163] Figure 8 The schematic diagram of the structure of the integrated pipe gallery risk identification system based on sound signal visualization according to an embodiment of the present application is as follows. The integrated pipe gallery risk identification system 800 based on sound signal visualization comprises a sound signal acquisition module 81, an image conversion module 82, a model training module 83, a risk assessment module 84, and a preprocessing module 85, wherein:

[0164] It should be noted that the sound signal acquisition module 71, the image conversion module 72, the model training module 73, the risk assessment module 74 and the sound signal acquisition module 81, the image conversion module 82, the model training module 83, the risk assessment module 84 have the same structure and function.

[0165] In an embodiment of the present application, the sound signal acquisition module 81 is also used to: collect real sound signals before and during the occurrence of various types of disasters in the integrated pipeline corridor; simulate the sounds before and during the occurrence of various types of disasters in the integrated pipeline corridor to generate simulated sound signals; and generate sample sound signals based on real sound signals and simulated sound signals.

[0166] In the embodiment of the present application, the preprocessing module 85 is also used to perform pre-emphasis processing operations, frame processing operations, windowing processing operations and noise reduction processing operations on real sound signals and simulated sound signals to generate sample sound signals.

[0167] In the embodiment of the present application, the image conversion module 82 is also used to: generate a virtual sound spectrogram based on the sound spectrogram of the sample sound signal; use the sound spectrogram and the virtual sound spectrogram as sample images, and annotate the sample images with detection boxes according to the disaster information corresponding to the sample sound signal.

[0168] In an embodiment of the present application, the risk assessment module 84 also includes: a terminal computing unit 841, which is used to obtain the attribute information of the target detection box on the sound spectrogram to be identified after disaster identification; and determine the average length, average height and average interval length of the target detection box based on the attribute information of each target detection box.

[0169] In an embodiment of the present application, the terminal computing unit 841 is also used to: obtain the size of the sound spectrogram to be identified after disaster identification; predict the average duration of the target sound signal when a disaster occurs based on the average length and size; obtain the frequency band range of the target sound signal based on the attribute information and size of the target detection frame; and obtain the average interval duration of adjacent target sound signals based on the average interval length and size.

[0170] In an embodiment of the present application, the risk assessment module 84 is also used to: input the sound spectrogram to be identified into the target disaster identification model to output the predicted detection box corresponding to the sound spectrogram to be identified; obtain the labeled detection box i that is missed by the target disaster identification model, and obtain the predicted detection box i-1 and predicted detection box i+1 adjacent to the labeled detection box i; based on the attribute information of the predicted detection box i-1 and the predicted detection box i+1, determine the attribute information of the predicted detection box i corresponding to the labeled detection box i; according to the attribute information of the predicted detection box i, output the disaster identification result in the sound sequence to be identified that is missed by the target disaster identification model.

[0171] In the embodiment of the present application, the network structures of the disaster recognition model, such as the feature extraction skeleton, detection layer, and fusion layer, are determined according to the feature complexity of the sample image and the size distribution of the detection frame annotation results.

[0172] The integrated pipe corridor risk identification system based on sound signal visualization proposed in this application collects real sound signals when disasters occur in the integrated pipe corridor and simulated sound signals in a laboratory simulation environment, and performs pre-processing such as pre-emphasis, windowing, framing and noise reduction on the collected real sound signals and simulated sound signals, thereby generating sample sound signals. The sample sound signal generates a corresponding sound spectrogram, and the sound spectrogram is determined as a sample image, and the sample image is annotated with a detection frame according to the target morphology and features of the sound signals of different types of disasters in the sound spectrogram. Based on the sample image and the annotated detection frame, the constructed disaster recognition model based on the deep learning target detection algorithm is trained, and the parameters of the disaster recognition model based on the deep learning target detection algorithm are adjusted based on the results of each round of training output until the training is completed and a trained target disaster recognition model is generated. Furthermore, based on the trained target disaster recognition model, the sound sequence to be recognized in the collected integrated pipe corridor is used for disaster recognition, and the recognition result output by the target disaster recognition model is corrected for missed detection. In this application, by collecting and analyzing the sound signals in the integrated pipeline corridor, the identification of possible disasters in the integrated pipeline corridor is achieved, the method of risk identification of the integrated pipeline corridor is improved, and the timeliness and accuracy of risk identification of the integrated pipeline corridor is effectively improved.

[0173] To implement the above embodiment, the present application also proposes an electronic device, such as Fig. 9 As shown, the electronic device 900 includes at least one processor 902; and a memory 901 communicatively connected to the at least one processor; wherein the memory 901 stores instructions executable by the at least one processor 902, and the instructions are executed by the at least one processor 902 so that the at least one processor 902 can execute the integrated pipeline corridor risk identification method based on sound signal visualization proposed in the above embodiment.

[0174] In order to implement the above embodiments, the present application also proposes a computer-readable storage medium, which stores a computer program. When the program is executed by a processor, the integrated pipeline corridor risk identification method based on sound signal visualization proposed in the above embodiments is implemented.

[0175] To implement the above embodiments, the present application also proposes a computer program product, including a computer program, which, when executed by a processor, implements the integrated pipeline corridor risk identification method based on sound signal visualization proposed in the above embodiments.

[0176] In the description of the present application, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", "axial", "radial", "circumferential" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present application.

[0177] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0178] In this application, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0179] In the present application, unless otherwise clearly specified and limited, a first feature being “above” or “below” a second feature may mean that the first and second features are in direct contact, or the first and second features are in indirect contact through an intermediate medium. Moreover, a first feature being “above”, “above”, and “above” a second feature may mean that the first feature is directly above or obliquely above the second feature, or simply means that the first feature is higher in level than the second feature. A first feature being “below”, “below”, and “below” a second feature may mean that the first feature is directly below or obliquely below the second feature, or simply means that the first feature is lower in level than the second feature.

[0180] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0181] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A method for identifying risks of integrated pipe gallery based on sound signal visualization, characterized in that: The following steps are involved: Collecting sound signals corresponding to multiple types of disasters inside the integrated pipe gallery as sample sound signals, wherein the sample sound signals are stored in the integrated pipe gallery sound information disaster database; Performing image processing on the sample sound signal to generate a sound spectrogram of the sample sound signal, using the sound spectrogram as a sample image of the sample sound signal, and marking the sample image with a detection frame based on disaster information corresponding to the sample sound signal; Based on the sample image and the corresponding annotated detection frame, the disaster recognition model constructed based on the deep learning target detection algorithm is trained to generate a trained target disaster recognition model; Collecting a sound sequence to be currently identified in the integrated pipe gallery, wherein the sound sequence includes a plurality of sound signals to be identified; Generating a sound spectrogram to be identified corresponding to the sound sequence to be identified, and inputting the sound spectrogram to be identified into the target disaster identification model to output a disaster identification result corresponding to the sound sequence to be identified; After the disaster is identified, the attribute information of the target detection box on the spectrogram of the sound to be identified is obtained; According to the attribute information of each of the target detection frames, the average length, average height and average interval length of the target detection frames are determined.

2. The method according to claim 1, characterized in that The collecting of sound signals corresponding to multiple types of disasters inside the integrated pipe gallery as sample sound signals includes: Collecting real sound signals before and during the occurrence of various disasters in the integrated pipeline corridor; Simulating the sounds before and during the occurrence of various disasters in the integrated pipe gallery to generate simulated sound signals; The sample sound signal is generated based on the real sound signal and the simulated sound signal.

3. The method according to claim 2, characterized in that The step of generating the sample sound signal based on the real sound signal and the simulated sound signal comprises: The real sound signal and the simulated sound signal are subjected to pre-emphasis processing, frame division processing, windowing processing and noise reduction processing to generate the sample sound signal.

4. The method according to claim 1, characterized in that: The method further comprises: Based on the sound spectrogram of the sample sound signal, creating a corresponding virtual sound spectrogram; The sound spectrogram and the virtual sound spectrogram are used as the sample image, and detection frames are marked on the sample image according to disaster information corresponding to the sample sound signal.

5. The method according to claim 1, characterized in that: The step of training the disaster recognition model constructed based on the deep learning target detection algorithm based on the sample image and the corresponding annotated detection frame to generate a trained target disaster recognition model includes: The network structures of the disaster recognition model based on the deep learning target detection algorithm, such as the feature extraction skeleton, detection layer, and fusion layer, are determined according to the feature complexity of the sample image and the size distribution of the detection frame annotation results.

6. The method according to claim 1, characterized in that The generating of the to-be-recognized sound spectrogram of the to-be-recognized sound sequence and inputting the to-be-recognized sound spectrogram into the target disaster identification model to output a disaster identification result corresponding to the to-be-recognized sound sequence includes: Inputting the to-be-recognized sound spectrogram into the target disaster recognition model to output a prediction detection box corresponding to the to-be-recognized sound spectrogram; Obtain a marked detection box i that is missed by the target disaster recognition model, and obtain a predicted detection box i-1 and a predicted detection box i+1 adjacent to the marked detection box i; Determine the attribute information of the predicted detection box i corresponding to the labeled detection box i based on the attribute information of the predicted detection box i-1 and the predicted detection box i+1; According to the attribute information of the predicted detection box i, the disaster recognition results in the sound sequence to be recognized that are missed by the target disaster recognition model are output.

7. The method according to claim 1, characterized in that The method further comprises: After the disaster is identified, the image size of the sound spectrogram to be identified is obtained; Predicting the average duration of the target sound signal when a disaster occurs based on the average length and the image size; Acquire a frequency band range where the target sound signal is located according to the attribute information of the target detection frame and the image size; The average interval duration of adjacent target sound signals is acquired according to the average interval length and the image size.

8. A comprehensive pipe gallery risk identification system based on sound signal visualization, characterized in that: It includes a sound signal acquisition module, a preprocessing module, an image conversion module, a model training module, and a risk assessment module. When the modules are running, the integrated pipeline corridor risk identification method based on sound signal visualization according to any one of claims 1-7 is implemented.

9. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to enable the computer to execute the method for risk identification of integrated pipe corridors based on sound signal visualization according to any one of claims 1 to 7.

Images