CN117668669B - Pipeline safety monitoring method and system based on improvement YOLOv (YOLOv) - Google Patents
Pipeline safety monitoring method and system based on improvement YOLOv (YOLOv) Download PDFInfo
- Publication number
- CN117668669B CN117668669B CN202410137294.6A CN202410137294A CN117668669B CN 117668669 B CN117668669 B CN 117668669B CN 202410137294 A CN202410137294 A CN 202410137294A CN 117668669 B CN117668669 B CN 117668669B
- Authority
- CN
- China
- Prior art keywords
- module
- yolov
- network
- dimensional space
- safety monitoring
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000012544 monitoring process Methods 0.000 title claims abstract description 76
- 238000000034 method Methods 0.000 title claims abstract description 60
- 230000006872 improvement Effects 0.000 title claims abstract description 22
- 238000012549 training Methods 0.000 claims abstract description 37
- 238000007781 pre-processing Methods 0.000 claims abstract description 10
- 238000012360 testing method Methods 0.000 claims abstract description 9
- 239000013307 optical fiber Substances 0.000 claims description 47
- 239000011159 matrix material Substances 0.000 claims description 41
- 238000001514 detection method Methods 0.000 claims description 38
- 230000006870 function Effects 0.000 claims description 15
- 230000008569 process Effects 0.000 claims description 15
- 230000009467 reduction Effects 0.000 claims description 14
- 238000006722 reduction reaction Methods 0.000 claims description 14
- 230000003287 optical effect Effects 0.000 claims description 13
- 238000010606 normalization Methods 0.000 claims description 9
- 238000002372 labelling Methods 0.000 claims description 8
- 238000005553 drilling Methods 0.000 claims description 7
- 238000000227 grinding Methods 0.000 claims description 7
- 238000012545 processing Methods 0.000 claims description 7
- 238000011161 development Methods 0.000 claims description 4
- 230000018109 developmental process Effects 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000008859 change Effects 0.000 claims description 3
- 238000007906 compression Methods 0.000 claims description 3
- 230000006835 compression Effects 0.000 claims description 3
- 238000013144 data compression Methods 0.000 claims description 3
- 238000004519 manufacturing process Methods 0.000 claims description 3
- 230000011218 segmentation Effects 0.000 claims description 3
- 238000012795 verification Methods 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 2
- 238000010586 diagram Methods 0.000 description 17
- 230000000694 effects Effects 0.000 description 9
- 238000013527 convolutional neural network Methods 0.000 description 7
- 239000000284 extract Substances 0.000 description 7
- 238000011176 pooling Methods 0.000 description 6
- 210000000988 bone and bone Anatomy 0.000 description 5
- 238000009412 basement excavation Methods 0.000 description 3
- 238000005070 sampling Methods 0.000 description 3
- 230000001427 coherent effect Effects 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 230000004807 localization Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
- G06F18/15—Statistical pre-processing, e.g. techniques for normalisation or restoring missing data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/217—Validation; Performance evaluation; Active pattern learning techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/2431—Multiple classes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/082—Learning methods modifying the architecture, e.g. adding, deleting or silencing nodes or connections
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
- G06F2218/04—Denoising
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Molecular Biology (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Probability & Statistics with Applications (AREA)
- Testing Or Calibration Of Command Recording Devices (AREA)
Abstract
The application discloses a pipeline safety monitoring method and system based on an improvement YOLOv (YOLOv), and relates to the technical field of pipeline safety monitoring. The application comprises the following steps: s1, acquiring two-dimensional space-time signals in a pipeline safety monitoring process, and preprocessing to obtain a two-dimensional space-time signal data set; s2, constructing an improved YOLOv network; s3, dividing a training set and a testing set based on the two-dimensional space-time signal data set, and training an improved YOLOv network by using the training set to obtain an improved YOLOv network model; s4, loading the two-dimensional space-time signal obtained in the pipeline safety monitoring process to the step; and S3, performing forward propagation once in the improved YOLOv network model, and obtaining the category and positioning information of the target event. The method can effectively improve the average accuracy of classifying and positioning the small target event.
Description
Technical Field
The invention relates to the technical field of pipeline safety monitoring, in particular to a pipeline safety monitoring method and system based on an improvement YOLOv (automatic control unit) 7.
Background
With the development of deep learning, many researchers use convolutional neural networks to classify target events in spatiotemporal data obtained based on distributed fiber vibration sensing systems (DVS). Because the convolutional neural network can automatically extract the characteristics of the target events in the space-time data, the convolutional neural network is used for classifying the target events in the space-time data in real application, and the method has stronger robustness. However, although the convolutional neural network can obtain a better classification result for the target event in the space-time data obtained based on the distributed optical fiber vibration sensing system (DVS), in terms of positioning the target event, when the convolutional neural network is used for target detection for the target event in the space-time data obtained based on the distributed optical fiber vibration sensing system (DVS), a traditional vibration intensity threshold positioning method is still used, and the traditional vibration intensity threshold positioning method is easily interfered by noise signals to cause event report missing.
Therefore, based on the requirements of classification tasks and positioning tasks of target events, researchers begin focusing on a target detection method using computer vision, such as a target detection method based on YOLOv target detectors, and the method has better target event classification results and higher target event positioning accuracy, can complete positioning and recognition tasks at the same time, has higher response speed and higher recognition positioning accuracy, can ensure the real-time performance of a system, and can eliminate noise and environmental interference influence through space-time logic, thereby effectively reducing false alarm rate and false alarm rate. However, when the data features are automatically extracted based on the YOLOv target detector in the prior art, there is a problem that part of feature information is lost, so that the target detection method based on the YOLOv target detector in the prior art still has a problem of low classification positioning accuracy when performing target detection on small target events (the small target events refer to that the target events have a short duration and are present in the whole long-distance spatiotemporal data) in the long-distance spatiotemporal data obtained based on the distributed optical fiber vibration sensing system (DVS).
Accordingly, applicants have devised an improved YOLOv-based pipeline security monitoring method and system that can improve the average accuracy of small target event localization and classification in long-range spatio-temporal data.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides a pipeline safety monitoring method and system based on an improvement YOLOv < 7 >.
In order to achieve the above purpose, the technical scheme of the invention is as follows:
a pipeline safety monitoring method based on improvement YOLOv, comprising the steps of:
s1, acquiring two-dimensional space-time signals in a pipeline safety monitoring process, and preprocessing to obtain a two-dimensional space-time signal data set;
S2, constructing an improved YOLOv network, wherein the improved YOLOv network refers to: the first SPD module and the second SPD module are used for respectively replacing a second CBS module and a fourth CBS module in the conventional YOLOv network; the third SPD module, the fourth SPD module and the fifth SPD module are used for respectively replacing a first MP module, a second MP module and a third MP module in a back bone part in the existing YOLOv network; adding a first CBAM module and a second CBAM module after a second ELAN module and a third ELAN module of the existing YOLOv network, wherein the first CBAM module is respectively connected with a third CBS module and a fourth SPD module, and the second CBAM module is respectively connected with the fourth CBS module and a fifth SPD module; the sixth SPD module and the seventh SPD module are used for respectively replacing a fourth MP module and a fifth MP module in the existing YOLOv network; the third CBAM module, the fourth CBAM module, the fifth CBAM module, the sixth CBAM module and the seventh CBAM module are added after the SPPCSPC module, the first Concat module, the second Concat module, the third Concat module and the fourth Concat module in the existing YOLOv network respectively, and the third CBAM module is connected with the fourth Concat module;
S3, dividing a training set and a testing set based on the two-dimensional space-time signal data set, and training an improved YOLOv network by using the training set to obtain an improved YOLOv network model;
And S4, loading the two-dimensional space-time signals obtained in the pipeline safety monitoring process into the improved YOLOv network model obtained in the step S3, and carrying out forward propagation once to obtain the category and positioning information of the target event.
Preferably, step S1 is specifically: and acquiring a two-dimensional space-time signal in the pipeline safety monitoring process, carrying out noise reduction, compression and normalization on the two-dimensional space-time signal to obtain a normalized data matrix, and then carrying out segmentation and labeling on the normalized data matrix to obtain a two-dimensional space-time signal data set, wherein the two-dimensional space-time signal data set is the two-dimensional space-time signal data set for pipeline safety monitoring.
Preferably, the step S1 specifically includes the following steps:
S101, constructing a pipeline safety monitoring signal acquisition system for acquiring two-dimensional space-time signals by utilizing a distributed vibration sensing system (DVS) based on a phase-sensitive photo-time domain reflectometer and a signal acquisition optical fiber;
S102, manually manufacturing electric drill drilling, sawing and grinding a pipeline, knocking the pipeline, walking by personnel, excavating and pipeline leakage for the pipeline provided with the signal acquisition optical fiber, and acquiring two-dimensional space-time signals of the six types of target events through a pipeline safety monitoring signal acquisition system;
s103, carrying out data noise reduction processing by utilizing two-dimensional space-time signal data to obtain a data matrix, then compressing the data matrix 256 times by adopting a differential average method, and carrying out normalization processing on the compressed data matrix to obtain a normalized data matrix;
S104, dividing the normalized data matrix into data matrix blocks with the height of 32 by taking 4S time (32 sampling points are covered in every 4S time) as a unit in a time domain;
S105, respectively labeling the data matrix blocks with the height of 32, wherein different types of events are respectively labeled corresponding to the events, then the labels are stored as npy format files, and the data matrix blocks with the height of 32 and the labels corresponding to the data matrix blocks respectively form a two-dimensional space-time signal data set of the data matrix type.
Preferably, the pipeline safety monitoring signal acquisition system comprises a distributed vibration sensing system (DVS) based on a phase sensitive photo time domain reflectometer and a signal acquisition optical fiber; the optical fiber distributed vibration sensing system (DVS) based on the phase-sensitive optical time domain reflectometer comprises an ultra-narrow linewidth laser, an acousto-optic modulator, an erbium-doped amplifier I, a circulator, an erbium-doped amplifier II, a photoelectric detector, a data acquisition card and an upper computer, wherein the ultra-narrow linewidth laser, the acousto-optic modulator, the erbium-doped amplifier I, the circulator, the erbium-doped amplifier II and the photoelectric detector are sequentially connected through sensing optical fibers, the photoelectric detector, the data acquisition card and the upper computer are sequentially connected through data wires, the circulator is further connected with a signal acquisition optical fiber, the signal acquisition optical fiber is a single-mode optical fiber, the signal acquisition optical fiber is paved along a pipeline as a detection optical cable, and the signal acquisition optical fiber is used for acquiring two-dimensional space-time signals for pipeline safety monitoring.
Preferably, the step S3 specifically includes the following steps:
S301, data matrix blocks with the height of 32 in the two-dimensional space-time signal data set and labels corresponding to the data matrix blocks are respectively processed according to 7:3, dividing the proportion to obtain a training set and a verification set;
S302, training the improved YOLOv network by using a training set under the guidance of a Loss function Loss of the existing YOLOv network to obtain an improved YOLOv network model.
Preferably, step S302 is specifically: the method comprises the steps of carrying out back propagation under the guidance of improved YOLOv network Loss obtained by Loss function Loss calculation of the existing YOLOv network, updating the weight of the improved YOLOv network, iterating for 50 times, and completing the training process of the improved YOLOv network to obtain an improved YOLOv network model; wherein the training process of the modified YOLOv network is based on the PyTorch development platform, and the training process of the modified YOLOv network, each training batch size is 64.
A pipeline safety monitoring system based on an improvement YOLOv comprises a two-dimensional space-time signal acquisition module, a two-dimensional space-time signal preprocessing module and a target event detection module; wherein,
A two-dimensional spatiotemporal signal acquisition module configured to: acquiring two-dimensional space-time signals in the pipeline safety monitoring process by utilizing a pipeline safety monitoring signal acquisition system;
A two-dimensional spatio-temporal signal preprocessing module configured to: carrying out data noise reduction, data compression and normalization on the two-dimensional space-time signal;
A target event detection module, having a modified YOLOv network model built in, and configured to: and inputting the two-dimensional space-time signals obtained in the pipeline safety monitoring process into an improved YOLOv network model to obtain the category and positioning information of the target event.
Compared with the prior art, the invention has the beneficial technical effects that:
The SPD module arranged in the YOLOv network can effectively avoid the loss of the characteristics in the process of extracting and downsampling the characteristics; the CBAM module arranged in the YOLOv network can enhance the characteristics of different channels and extract key information of different positions in space, so that the effect of learning and extracting the characteristic information of all target events including small target events by the YOLOv network is effectively improved, the YOLOv network is improved to better capture global context information of all target events including small target events, the loss of the characteristic information and the global context information of the small target events is effectively avoided, and the aim of improving the average accuracy of classifying and positioning the small target events is fulfilled.
Drawings
FIG. 1 is a flow chart of a pipeline safety monitoring method based on improvement YOLOv of the present application;
FIG. 2 is a schematic diagram of connection relation of a pipeline safety monitoring signal acquisition system according to the present application;
FIG. 3 is a two-dimensional space-time signal of a drill hole target event acquired by a pipeline safety monitoring signal acquisition system, wherein the abscissa represents a distance in m, the distance is the distance between a signal acquisition position of the target event and a connection position of a circulator and a signal acquisition optical fiber, and the ordinate represents acquisition time of the target event in s;
FIG. 4 is a two-dimensional space-time signal of a sawmilling pipeline target event acquired by a pipeline safety monitoring signal acquisition system, wherein the abscissa represents a distance in m, the distance is the distance between a signal acquisition position of the target event and a connection position of a circulator and a signal acquisition optical fiber, and the ordinate represents acquisition time of the target event in s;
FIG. 5 is a two-dimensional space-time signal of a knocking pipeline target event obtained by a pipeline safety monitoring signal acquisition system, wherein the abscissa represents a distance in m, the distance is the distance between a signal acquisition position of the target event and a connection position of a circulator and a signal acquisition optical fiber, and the ordinate represents acquisition time of the target event in s;
FIG. 6 is a two-dimensional space-time signal of a personnel walking target event acquired by a pipeline safety monitoring signal acquisition system, wherein the abscissa represents a distance in m, the distance is the distance between a signal acquisition position of the target event and a connection position of a circulator and a signal acquisition optical fiber, and the ordinate represents acquisition time of the target event in s;
FIG. 7 is a two-dimensional spatio-temporal signal of an excavation target event acquired by a pipeline safety monitoring signal acquisition system, wherein the abscissa represents a distance in m, the distance is the distance between a signal acquisition position of the target event and a connection position of a circulator and a signal acquisition optical fiber, and the ordinate represents acquisition time of the target event in s;
FIG. 8 is a two-dimensional space-time signal of a pipeline leakage target event acquired by a pipeline safety monitoring signal acquisition system, wherein the abscissa represents a distance in m, the distance is the distance between a signal acquisition position of the target event and a connection position of a circulator and a signal acquisition optical fiber, and the ordinate represents acquisition time of the target event in s;
FIG. 9 is a schematic diagram of the structure of an SPD module of the present application;
FIG. 10 is a block diagram of a modified YOLOv network according to the present application;
fig. 11 is a block diagram of a conventional YOLOv network according to the present application.
Detailed Description
As shown in fig. 1, the embodiment discloses a pipeline safety monitoring method based on improvement YOLOv7, which comprises the following steps:
s1, acquiring two-dimensional space-time signals in a pipeline safety monitoring process, and preprocessing to obtain a two-dimensional space-time signal data set;
Specifically, two-dimensional space-time signals in the pipeline safety monitoring process are collected, noise reduction, compression and normalization are carried out on the two-dimensional space-time signals to obtain a normalized data matrix, and then segmentation and labeling are carried out on the normalized data matrix to obtain a two-dimensional space-time signal data set, wherein the two-dimensional space-time signal data set is the two-dimensional space-time signal data set for pipeline safety monitoring;
The step S1 specifically includes the following steps:
S101, constructing a pipeline safety monitoring signal acquisition system for acquiring two-dimensional space-time signals by utilizing a distributed vibration sensing system (DVS) based on a phase-sensitive photo-time domain reflectometer and a signal acquisition optical fiber;
As shown in fig. 2, the pipeline safety monitoring signal acquisition system comprises a distributed vibration sensing system (DVS) based on a phase sensitive photo time domain reflectometer and a signal acquisition optical fiber; the optical fiber distributed vibration sensing system (DVS) based on the phase sensitive optical time domain reflectometer is in the prior art, the optical fiber distributed vibration sensing system (DVS) based on the phase sensitive optical time domain reflectometer comprises an ultra-narrow linewidth laser, an acousto-optic modulator, an erbium-doped amplifier I, a circulator, an erbium-doped amplifier II, a photoelectric detector, a data acquisition card and an upper computer, wherein the ultra-narrow linewidth laser, the acousto-optic modulator, the erbium-doped amplifier I, the circulator, the erbium-doped amplifier II and the photoelectric detector are sequentially connected through sensing optical fibers, the photoelectric detector, the data acquisition card and the upper computer are sequentially connected through data lines, the circulator is also connected with a signal acquisition optical fiber, the signal acquisition optical fiber is a single-mode optical fiber, the signal acquisition optical fiber is paved along a pipeline as a detection optical cable, the signal acquisition optical fiber is used for acquiring two-dimensional space-time signals for pipeline safety monitoring, and grinding pipeline safety monitoring signals comprise two-dimensional signals of drilling of an electric drill, a pipeline, a staff walking, a digging and pipeline leakage event;
The principle of the pipeline safety monitoring signal acquisition system for acquiring two-dimensional space-time signals is as follows: the principle of the pipeline safety monitoring signal acquisition system for acquiring two-dimensional space-time signals is as follows: an ultra-narrow linewidth laser with linewidth of 3kHz is used as a light source, the ultra-narrow linewidth laser is used as a laser source to generate a continuous coherent light signal, the continuous coherent light signal is modulated into an optical pulse signal through an acousto-optic modulator, the optical pulse signal is amplified by an erbium-doped optical fiber amplifier I, the amplified optical pulse signal is transmitted to a signal acquisition optical fiber through a port 1 and a port 2 of a circulator, the optical pulse signal generates a backward Rayleigh scattering light signal in the transmission process of the signal acquisition optical fiber, the backward Rayleigh scattering light signal returns to the port 2 of the circulator along the signal acquisition optical fiber, the backward Rayleigh scattering light signal is transmitted to an erbium-doped amplifier II through the port 3 of the circulator to be amplified, the amplified backward Rayleigh scattering light signal is transmitted to a photoelectric detector, the photoelectric detector converts the received backward Rayleigh scattering light signal into an electric signal, a data acquisition card acquires the electric signal, and the acquired electric signal is transmitted to an upper computer, and the data acquisition card phase demodulates and upper computer software processes the electric signal to acquire an action signal vibrated on the signal acquisition optical fiber.
S102, manually manufacturing electric drill drilling, sawing and grinding a pipeline, knocking the pipeline, walking by personnel, excavating and pipeline leakage for the pipeline provided with the signal acquisition optical fiber, and acquiring two-dimensional space-time signals of the six types of target events through a pipeline safety monitoring signal acquisition system; the two-dimensional space-time signals of six target events, namely electric drill drilling, sawing and grinding the pipeline, knocking the pipeline, walking by personnel, excavating and pipeline leakage, acquired by the pipeline safety monitoring signal acquisition system are respectively shown in fig. 3 to 8.
S103, carrying out data noise reduction processing by utilizing two-dimensional space-time signal data to obtain a data matrix, then compressing the data matrix 256 times by adopting a differential average method, and carrying out normalization processing on the compressed data matrix to obtain a normalized data matrix;
S104, dividing the normalized data matrix into data matrix blocks with the height of 32 by taking 4S time (32 sampling data are covered in every 4S time) as a unit in a time domain; the time length of 4s in the present application is sufficient to contain the complete vibration signal of a target event;
s105, respectively labeling the data matrix blocks with the height of 32, respectively labeling different types of events corresponding to the events, then storing the labels as npy format files, and respectively labeling the data matrix blocks with the height of 32 to form a two-dimensional space-time signal data set of the data matrix type.
S2, constructing an improved YOLOv network:
Modified YOLOv network as shown in fig. 10, the modified YOLOv network of the present application includes an Input part, a modified Backbone part, and a modified Head part connected in sequence;
the Input part comprises an Input layer;
the improved backhaul part comprises a first CBS module, a first SPD module, a second CBS module, a second SPD module, a first ELAN module, a third SPD module, a second ELAN module, a first CBAM module, a fourth SPD module, a third ELAN module, a second CBAM module, a fifth SPD module and a fourth ELAN module which are connected in sequence; the backup part in the improved YOLOv network is used for extracting characteristic information;
The improved Head part comprises SPPCSPC modules, a third CBAM module, a fifth CBS module, a first UP module, a first Concat module, a fourth CBAM module, a fifth ELAN module, a sixth CBS module, a second UP module, a second Concat module, a fifth CBAM module, a sixth ELAN module, a sixth SPD module, a third Concat module, a sixth CBAM module, a seventh ELAN module, a seventh SPD module, a fourth Concat module, a seventh CBAM module, an eighth ELAN module, a third REP module and a third CONV module which are sequentially connected; the sixth ELAN module is further connected with a first REP module and a first CONV module in sequence, and the seventh ELAN module is further connected with a second REP module and a second CONV module in sequence; the first CBAM module is also connected with a third CBS module and a second Concat module in sequence, and the second CBAM module is connected with a fourth CBS module and a first Concat module in sequence; the third CBAM module is also connected to the fourth Concat module, and the fifth ELAN module is also connected to the third Concat module;
In the present application, in step S2, the modified YOLOv network (as shown in fig. 10) is obtained by modifying the existing YOLOv network (as shown in fig. 11) as follows, specifically:
the first SPD module is utilized, and the second SPD module is utilized to replace a second CBS module and a fourth CBS module in a backhaul part in the existing YOLOv network respectively;
the third SPD module, the fourth SPD module and the fifth SPD module are used for respectively replacing a first MP module, a second MP module and a third MP module in a back bone part in the existing YOLOv network;
Adding a first CBAM module and a second CBAM module after a second ELAN module and a third ELAN module in a Backbone portion of the existing YOLOv network, wherein the first CBAM module is connected with a third CBS module and a fourth SPD module, and the second CBAM module is connected with the fourth CBS module and a fifth SPD module, respectively;
replacing a fourth MP module and a fifth MP module in the Head part of the existing YOLOv network by a sixth SPD module and a seventh SPD module respectively;
the third CBAM module, the fourth CBAM module, the fifth CBAM module, the sixth CBAM module and the seventh CBAM module are added after the SPPCSPC module, the first Concat module, the second Concat module, the third Concat module and the fourth Concat module in the Head portion of the existing YOLOv7 network, respectively, and the third CBAM module is connected to the fourth Concat module, and the fifth ELAN module is connected to the third Concat module.
In addition to the improvement on the existing YOLOv network, other modules in the existing YOLOv network and functions of the modules thereof are adopted in the improved YOLOv network correspondingly, so that the functions of the existing modules in the existing YOLOv network in the improved YOLOv7 network are adopted, and the application is not repeated;
The functions of each module for improving the existing YOLOv network in the application are as follows:
The structures and the functions of the first SPD module to the seventh SPD module are the same, and the functions of the first SPD module to the seventh SPD module are all extracted features and downsampled; all the characteristic information can be reserved to the maximum degree in the downsampling process of the first SPD module to the seventh SPD module, so that the characteristic information loss of a small target event is avoided, and the characteristic information loss is mainly caused by the following steps:
All the SPD module structures adopted in the improved YOLOv network in the present application are identical and are consistent with the SPD module structure disclosed in paper No More Strided Convolutions or Pooling: A New CNN Building Block for Low-Resolution Images and Small Objects; all SPD module structures in the YOLOv network modified in the application are shown in figure 9, and are composed of a space-to-depth layer and a non-stride convolution layer connected with the space-to-depth layer; the working principle of all SPD modules in the YOLOv network improved by the application is that a space-to-depth layer is firstly utilized to cut a large feature map into a plurality of small feature maps with the same size, then the space-to-depth layer is used to splice the small feature maps, and then a non-stride convolution layer (the step length is 1) is used to change the channel number of the spliced feature maps, so as to obtain the feature map after downsampling. In this embodiment, in the process of cutting the large feature map into several small feature maps with the same size by using the space-to-depth layer, the downsampled scale factor is set to 2. In the application, the step sizes of a second CBS module and a fourth CBS module in a back bone part in the existing YOLOv network are 2, the step sizes of the other CBS modules are 1, and when in downsampling, the step sizes of the second CBS module and the fourth CBS module are larger than 1, the moving distance of a convolution kernel is far, which can lead to the reduction of the spatial resolution of an output characteristic diagram, the reduction of the sampling position of the convolution kernel can also lead to the problem of characteristic loss, and the step sizes of the second CBS module and the fourth CBS module in the existing YOLOv network are 2, and can lead to the reduction of the receptive field of the convolution kernel (the receptive field refers to the size of the area of each output element in the convolution layer affected by an input element), the quantity of the input element in the receptive field of the convolution kernel is reduced, and the perceived context information is also reduced, thereby reducing the richness and the expression capability of the characteristics; in addition, the five MP modules (i.e., the first MP module to the fifth MP module) in the existing YOLOv network all include the largest pooling layer, the effect is downsampling, and in the process of adopting the pooling operation of the largest pooling layer, the input element within the range of each pooling window can only extract one maximum value as output, therefore, the operation can lead to the reduction of the spatial resolution of the input feature map, the loss of part of fine granularity features, and the reduction of the receptive field, and the receptive field gradually reduces in space along with the stacking of the pooled layers, so that the capability of the model in processing global information is reduced; instead of downsampling by using convolution with step length of 2 and a maximum pooling layer, the SPD module adopted in the application divides the feature map into a series of small feature maps and then splices the feature maps, and then adopts non-stride convolution to change channels, so that loss of the characterizable can be avoided, and all characteristic information is reserved to the maximum extent.
All CBAM modules in the YOLOv network improved by the application have the same structure, and CBAM modules disclosed in the paper publication CBAM: convolutional Block Attention Module are adopted. The first CBAM module, the second CBAM module and the third CBAM module in the improved YOLOv network of the present application are respectively disposed at three branches led out by the backhaul portion, which enables the improved YOLOv network to better learn and extract feature information of all target events including small target events and better capture global context information of all target events including small target events, specifically:
The first CBAM module is connected with the third CBS module of the first branch led out by the back bone part, the channel attention of the first CBAM module enhances the characteristics of different channels of the characteristic diagram output by the second ELAN module, the space attention of the first CBAM module extracts key information at different positions in the space of the characteristic diagram output by the second ELAN module to obtain a characteristic diagram with enhanced characteristic information and position information, and then the first CBAM module transmits the characteristic diagram with enhanced characteristic information and position information to the fourth SPD module for extracting characteristics and downsampling and simultaneously transmits the extracted characteristic diagram to the third CBS module for splicing, so that the whole improved YOLOv7 network can more fully master global context information of all target events including small target events;
The second CBAM module is connected with the fourth CBS module of the second branch led out by the back bone part, the channel attention of the second CBAM module enhances the characteristics of different channels of the characteristic diagram output by the second ELAN module, the space attention of the first CBAM module extracts key information at different positions in the space of the characteristic diagram output by the third ELAN module to obtain a characteristic diagram with enhanced characteristic information and position information, and then the second CBAM module transmits the characteristic diagram with enhanced characteristic information and position information to the fifth SPD module for extracting characteristics and downsampling and simultaneously transmits the extracted characteristic diagram to the fourth CBS module for splicing, so that the whole improved YOLOv7 network can more fully master global context information of all target events including small target events;
The third CBAM module is connected with the SPPCSPC module of the third branch led out by the backbond part, the channel attention of the third CBAM module enhances the different channel characteristics of the feature map which is output by the SPPCSPC module and is added with the receptive field, the space attention of the third CBAM module extracts key information at different positions in the space of the feature map which is output by the SPPCSPC module and is added with the receptive field to obtain a feature information and position information enhanced feature map, and then the third CBAM module transmits the feature information and position information enhanced feature map to the fifth CBS module for feature extraction and simultaneously transmits the feature information and the position information enhanced feature map to the fourth Concat module for splicing, so that the whole improved YOLOv network can more fully grasp global context information of all target events including small target events;
in the improved YOLOv network, after the first Concat module, the second Concat module, the third Concat module and the fourth Concat module of the Head part are respectively added with the fourth CBAM module, the fifth CBAM module, the sixth CBAM module and the seventh CBAM module, the functions of the fourth CBAM module to the seventh CBAM module are the same, the different channel characteristics of the characteristic diagrams output by the Concat module adjacent to the fourth YOLOv module are enhanced, and the key information of different positions in the space of the characteristic diagrams output by the Concat module adjacent to the fourth CBAM module are extracted, so that the recognition precision of small target events is effectively improved.
S3, dividing a training set and a testing set based on the two-dimensional space-time signal data set, and training an improved YOLOv network by using the training set to obtain an improved YOLOv network model;
the step S3 specifically includes the following steps:
S301, data matrix blocks with the height of 32 in the two-dimensional space-time signal data set and labels corresponding to the data matrix blocks are respectively processed according to 7:3, dividing the proportion to obtain a training set and a verification set;
S302, training an improved YOLOv network by using a training set under the guidance of a Loss function Loss of the existing YOLOv network to obtain an improved YOLOv network model;
Specifically, the method comprises the steps of carrying out back propagation under the guidance of improved YOLOv network Loss obtained by Loss function Loss calculation of the existing YOLOv network, updating the weight of the improved YOLOv network, iterating for 50 times, and completing the training process of the improved YOLOv network to obtain an improved YOLOv7 network model; wherein the training process of the modified YOLOv network is based on the PyTorch development platform, and the training process of the modified YOLOv network, each training batch size is 64.
The Loss function Loss of the existing YOLOv network is the sum of category confidence Loss (f a), coordinate regression Loss (f b) and target confidence Loss (f c), and is shown in formula (1);
Loss=a1* fa+a2* fb+ a3* fc （1）
a 1、a2、a3 is a weight coefficient, wherein a 1 is 0.125, a 2 is 0.05, and a 3 is 0.1;
And S4, loading the two-dimensional space-time signals obtained in the pipeline safety monitoring process into the improved YOLOv network model obtained in the step S3, and carrying out forward propagation once to obtain the category and positioning information of the target event.
A pipeline safety monitoring system based on improvement YOLOv comprises a two-dimensional space-time signal acquisition module, a two-dimensional space-time signal preprocessing module and a target event detection module, wherein,
A two-dimensional spatiotemporal signal acquisition module configured to: acquiring two-dimensional space-time signals in the pipeline safety monitoring process by utilizing a pipeline safety monitoring signal acquisition system;
A two-dimensional spatio-temporal signal preprocessing module configured to: carrying out data noise reduction, data compression and normalization on the two-dimensional space-time signal;
A target event detection module, built-in based on the modified YOLOv network model, and configured to: and inputting the two-dimensional space-time signals obtained in the pipeline safety monitoring process into an improved YOLOv network model to obtain the category and positioning information of the target event.
In order to verify the effect of the improved YOLOv 7-based pipeline security monitoring method in terms of small target event classification and positioning, the application also compares the detection effect of the improved YOLOv 7-based pipeline security monitoring method (abbreviated as YOLOv7-SPD-CBAM in Table 1), the Faster-RCNN target detection method (from Faster R-CNN: towards Real-Time Object Detection with Region Proposal Networks), the YOLOv5 target detection method (from https:// gitub. Com/ultralytics/yolov 5), the YOLOv7 target detection method (from YOLOv: trainer bag-of-freebies SETS NEW STATE-of-the-art for real-time object detectors), the YOLOv network model in the method is shown in FIG. 11), and in order to ensure fairness of the test, the application performs the same strategy as the above-mentioned 3 existing target detection methods (namely Faster-RCNN target detection method, YOLOv target detection method and YOLOv target detection method) and the improved pipeline security detection method based on the same training strategy as the application training method in Table 1, and then performs the same test strategy as the above-mentioned method based on the training strategy of the application, and the application is adopted for testing the same test strategy as the training method shown in the application, and the application is shown in the test model 1, and the application is better.
TABLE 1
| Target detection method | Precision | Recall | F1 | mAP@.5 | Obejctness |
| Faster-RCNN | 0.934 | 0.948 | 0.941 | 94.3% | - |
| YOLOv5 | 0.968 | 0.958 | 0.963 | 96.2% | 0.63% |
| YOLOv7 | 0.978 | 0.970 | 0.973 | 97.8% | 0.47% |
| YOLOv7-SPD-CBAM | 0.994 | 0.992 | 0.993 | 99.7% | 0.22% |
In table 1, precision, recall, F, map@5, and Obejctness were used as evaluation indexes to evaluate the detection effect of each network model in the above four methods.
Precision represents the accuracy of each network model detection in the above four methods, that is, the proportion of the samples with the prediction results being positive examples that are actually positive samples; recall represents Recall rate of each network model detection in the four methods, namely, the prediction result is that the actual positive sample number in the positive samples accounts for the proportion of the positive samples in the whole samples; f1 represents the reconciliation average of accuracy Precision and Recall rate Recall of each network model detection in the four methods; mAP@.5 represents the average accuracy of detection of each network model in the four methods when IOU=0.5; obejctness denotes the target detection loss average. The larger Precision, recall, F and mAP@5 are, the better; the smaller the Obejctness number, the better the target detection effect.
The pipeline safety monitoring method based on the improvement YOLOv7 has the best effect on four evaluation indexes of Precision, recall, F and mAP@5. Because YOLOv network model effect is best in the three existing target detection methods, the application focuses on comparing the improved YOLOv network model with YOLOv network model:
Precision index comparison: compared with the YOLOv target detection method in the prior art, the pipeline safety monitoring method based on the improvement YOLOv7 reaches 0.994 on the Precision index, and is improved by (0.994-0.978)/0.978 multiplied by 100% = 1.6%;
Recall index contrast: compared with the YOLOv7 target detection method in the prior art, the pipeline safety monitoring method based on the improvement YOLOv7 reaches 0.992 on the Recall index, and is improved by (0.992-0.970)/0.970 multiplied by 100% = 2.3%;
f1 index comparison: compared with the YOLOv7 target detection method in the prior art, the pipeline safety monitoring method based on the improvement YOLOv7 reaches 0.993 on the F1 index, and is improved by (0.993-0.973)/0.973 multiplied by 100% = 2.1%;
mAP@.5 index contrast: compared with the YOLOv target detection method in the prior art, the pipeline safety monitoring method based on the improvement YOLOv7 reaches 0.997 on mAP@5 index, and is improved by (0.997-0.978)/0.978 multiplied by 100% = 1.9%; the mAP@5 index is better, which indicates that the improved YOLOv-based pipeline safety monitoring method is better in average accuracy in classification and positioning when six small target events such as electric drill drilling, pipeline sawing and grinding, pipeline knocking, personnel walking, excavation and pipeline leakage are detected.
Objectness index contrast: compared with the YOLOv7 target detection method in the prior art, the pipeline safety monitoring method based on the improvement YOLOv7 achieves 0.22% on Objectness index, and is reduced by (0.47% -0.22%)/0.47%. Times.100% = 53.19%; the index Objectness in the application is better, which shows that the pipeline safety monitoring method based on the improvement YOLOv is more accurate when detecting six types of small target events such as electric drill drilling, pipeline sawing and grinding, pipeline knocking, personnel walking, excavation and pipeline leakage.
While the foregoing description of the embodiments of the present invention has been presented in conjunction with the drawings, it should be understood that it is not intended to limit the scope of the invention, but rather, it is intended to cover all modifications or variations within the scope of the invention as defined by the claims of the present invention.
Claims (7)
1. The pipeline safety monitoring method based on the improvement YOLOv is characterized by comprising the following steps of: the method comprises the following steps:
s1, acquiring two-dimensional space-time signals in a pipeline safety monitoring process, and preprocessing to obtain a two-dimensional space-time signal data set;
s2, constructing an improved YOLOv network;
The improved YOLOv network refers to: the first SPD module and the second SPD module are used for respectively replacing a second CBS module and a fourth CBS module in the existing YOLOv network; replacing the first MP module to the third MP module in the existing YOLOv network by the third SPD module to the fifth SPD module respectively; adding a first CBAM module and a second CBAM module after a second ELAN module and a third ELAN module in the existing YOLOv network, wherein the first CBAM module is respectively connected with the third CBS module and the fourth SPD module, and the second CBAM module is respectively connected with the fourth CBS module and the fifth SPD module; the sixth SPD module and the seventh SPD module are used for respectively replacing a fourth MP module and a fifth MP module in the existing YOLOv network; adding a third CBAM module to a seventh CBAM module after the SPPCSPC module and the first Concat module to the fourth Concat module in the existing YOLOv network, respectively, and connecting the third CBAM module to the fourth Concat module;
S3, dividing a training set and a testing set based on the two-dimensional space-time signal data set, and training an improved YOLOv network by using the training set to obtain an improved YOLOv network model;
S4, loading the two-dimensional space-time signals obtained in the pipeline safety monitoring process into an improved YOLOv network model for forward propagation once, and obtaining the category and positioning information of the target event;
the structures and the functions of the first SPD module and the seventh SPD module are the same, the functions of the first SPD module and the seventh SPD module are the feature extraction and downsampling, and all the SPD module structures are composed of a space-to-depth layer and a non-stride convolution layer connected with the space-to-depth layer; all SPD modules firstly cut a large feature map into a plurality of small feature maps with the same size by using a space-to-depth layer, then splice the small feature maps, and then change the channel number of the spliced feature maps by using a non-stride convolution layer to obtain a feature map after downsampling, wherein the step length of the non-stride convolution layer is 1; in the process of cutting a large feature map into a plurality of small feature maps with the same size by using a space-to-depth layer, the downsampled scale factor is set to be 2.
2. The improved YOLOv-based pipeline safety monitoring method as set forth in claim 1, wherein: the step S1 specifically comprises the following steps: and acquiring a two-dimensional space-time signal in the pipeline safety monitoring process, carrying out noise reduction, compression and normalization on the two-dimensional space-time signal to obtain a normalized data matrix, and then carrying out segmentation and labeling on the normalized data matrix to obtain a two-dimensional space-time signal data set, wherein the two-dimensional space-time signal data set is the two-dimensional space-time signal data set for pipeline safety monitoring.
3. A method of pipeline safety monitoring based on the improvement YOLOv as defined in claim 2, wherein: the step S1 specifically comprises the following steps:
S101, constructing a pipeline safety monitoring signal acquisition system for acquiring two-dimensional space-time signals by utilizing an optical fiber distributed vibration sensing system and a signal acquisition optical fiber based on a phase-sensitive optical time domain reflectometer;
S102, manually manufacturing electric drill drilling, sawing and grinding a pipeline, knocking the pipeline, walking by personnel, excavating and pipeline leakage for the pipeline provided with the signal acquisition optical fiber, and acquiring two-dimensional space-time signals of the six types of target events through a pipeline safety monitoring signal acquisition system;
s103, carrying out data noise reduction processing by utilizing two-dimensional space-time signal data to obtain a data matrix, then compressing the data matrix 256 times by adopting a differential average method, and carrying out normalization processing on the compressed data matrix to obtain a normalized data matrix;
s104, dividing the normalized data matrix in a time domain by taking 4S time as a unit to obtain a data matrix block with the height of 32;
S105, respectively labeling the data matrix blocks with the height of 32, wherein the events of different categories are respectively labeled corresponding to the events, then storing the labels as npy format files, and the data matrix blocks with the height of 32 and the labels corresponding to the data matrix blocks respectively form a two-dimensional space-time signal data set of the data matrix type.
4. A method of pipeline safety monitoring based on the improvement YOLOv as set forth in claim 3, wherein: the system for collecting the pipeline safety monitoring signals comprises an optical fiber distributed vibration sensing system and a signal collecting optical fiber based on a phase sensitive optical time domain reflectometer; the optical fiber distributed vibration sensing system based on the phase-sensitive optical time domain reflectometer comprises an ultra-narrow linewidth laser, an acousto-optic modulator, an erbium-doped amplifier I, a circulator, an erbium-doped amplifier II, a photoelectric detector, a data acquisition card and an upper computer, wherein the ultra-narrow linewidth laser, the acousto-optic modulator, the erbium-doped amplifier I, the circulator, the erbium-doped amplifier II and the photoelectric detector are sequentially connected through sensing optical fibers, the photoelectric detector, the data acquisition card and the upper computer are sequentially connected through data lines, the circulator is further connected with a signal acquisition optical fiber, the signal acquisition optical fiber is a single-mode optical fiber, the signal acquisition optical fiber is paved along a pipeline as a detection optical cable, and the signal acquisition optical fiber is used for acquiring two-dimensional space-time signals for pipeline safety monitoring.
5. The improved YOLOv-based pipeline safety monitoring method as set forth in claim 1, wherein: the step S3 specifically comprises the following steps:
S301, data matrix blocks with the height of 32 in the two-dimensional space-time signal data set and labels corresponding to the data matrix blocks are respectively processed according to 7:3, dividing the proportion to obtain a training set and a verification set;
S302, training the improved YOLOv network by using a training set under the guidance of a Loss function Loss of the existing YOLOv network to obtain an improved YOLOv network model.
6. The improved YOLOv-based pipeline safety monitoring method as set forth in claim 5, wherein: the step S302 specifically includes: the method comprises the steps of carrying out back propagation under the guidance of improved YOLOv network Loss obtained by Loss function Loss calculation of the existing YOLOv network, updating the weight of the improved YOLOv network, iterating for 50 times, and completing the training process of the improved YOLOv network to obtain an improved YOLOv network model; wherein the training process of the modified YOLOv network is based on the PyTorch development platform, and the training process of the modified YOLOv network, each training batch size is 64.
7. A pipeline safety monitoring system based on improvement YOLOv, characterized in that: the system comprises a two-dimensional space-time signal acquisition module, a two-dimensional space-time signal preprocessing module and a target event detection module; wherein,
A two-dimensional spatiotemporal signal acquisition module configured to: acquiring two-dimensional space-time signals in the pipeline safety monitoring process by utilizing a pipeline safety monitoring signal acquisition system;
A two-dimensional spatio-temporal signal preprocessing module configured to: carrying out data noise reduction, data compression and normalization on the two-dimensional space-time signal;
a target event detection module having the improved YOLOv network model of any one of claims 1 to 6 built-in and configured to: and inputting the two-dimensional space-time signals obtained in the pipeline safety monitoring process into an improved YOLOv network model to obtain the category and positioning information of the target event.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410137294.6A CN117668669B (en) | 2024-02-01 | 2024-02-01 | Pipeline safety monitoring method and system based on improvement YOLOv (YOLOv) |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410137294.6A CN117668669B (en) | 2024-02-01 | 2024-02-01 | Pipeline safety monitoring method and system based on improvement YOLOv (YOLOv) |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117668669A CN117668669A (en) | 2024-03-08 |
| CN117668669B true CN117668669B (en) | 2024-04-19 |
Family
ID=90084752
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410137294.6A Active CN117668669B (en) | 2024-02-01 | 2024-02-01 | Pipeline safety monitoring method and system based on improvement YOLOv (YOLOv) |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117668669B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118587560A (en) * | 2024-07-31 | 2024-09-03 | 齐鲁工业大学(山东省科学院) | Optical fiber distributed low signal-to-noise ratio target monitoring method and system |
Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114220169A (en) * | 2021-12-16 | 2022-03-22 | 山东大学 | Lightweight real-time monitoring abnormal behavior detection method based on Yolo-TSM |
| CN115223009A (en) * | 2022-07-07 | 2022-10-21 | 云南电网有限责任公司电力科学研究院 | Small target detection method and device based on improved YOLOv5 |
| CN115376028A (en) * | 2022-08-24 | 2022-11-22 | 四川九洲空管科技有限责任公司 | Target detection method based on dense feature point splicing and improved YOLOV5 |
| CN115471670A (en) * | 2022-07-20 | 2022-12-13 | 中国科学院西安光学精密机械研究所 | Space target detection method based on improved YOLOX network model |
| CN115731533A (en) * | 2022-11-29 | 2023-03-03 | 淮阴工学院 | Vehicle-mounted target detection method based on improved YOLOv5 |
| CN115964653A (en) * | 2022-12-27 | 2023-04-14 | 泰山学院 | Underwater sound signal classification method, system, equipment and storage medium |
| CN115984698A (en) * | 2023-01-05 | 2023-04-18 | 华南农业大学 | Litchi fruit growing period identification method based on improved YOLOv5 |
| CN116129329A (en) * | 2023-03-02 | 2023-05-16 | 南京易视萃智科技有限公司 | Early-stage tiny flame detection method based on deep-space deconvolution and Kalman filtering |
| WO2023103372A1 (en) * | 2021-12-06 | 2023-06-15 | 江苏理工学院 | Recognition method in state of wearing mask on human face |
| CN116580324A (en) * | 2023-05-08 | 2023-08-11 | 长春理工大学 | Yolov 5-based unmanned aerial vehicle ground target detection method |
| CN116645328A (en) * | 2023-04-25 | 2023-08-25 | 浙江理工大学 | Intelligent detection method for surface defects of high-precision bearing ring |
| WO2023185243A1 (en) * | 2022-03-29 | 2023-10-05 | 河南工业大学 | Expression recognition method based on attention-modulated contextual spatial information |
| CN117115686A (en) * | 2023-07-27 | 2023-11-24 | 中南民族大学 | Urban low-altitude small unmanned aerial vehicle detection method and system based on improved YOLOv7 |
| WO2024000709A1 (en) * | 2022-06-30 | 2024-01-04 | 西安交通大学 | Automatic seismic facies identification method based on combination of self-attention mechanism and u-shaped structure |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111598866B (en) * | 2020-05-14 | 2023-04-11 | 四川大学 | Lens key feature positioning method based on eye B-ultrasonic image |
-
2024
- 2024-02-01 CN CN202410137294.6A patent/CN117668669B/en active Active
Patent Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2023103372A1 (en) * | 2021-12-06 | 2023-06-15 | 江苏理工学院 | Recognition method in state of wearing mask on human face |
| CN114220169A (en) * | 2021-12-16 | 2022-03-22 | 山东大学 | Lightweight real-time monitoring abnormal behavior detection method based on Yolo-TSM |
| WO2023185243A1 (en) * | 2022-03-29 | 2023-10-05 | 河南工业大学 | Expression recognition method based on attention-modulated contextual spatial information |
| WO2024000709A1 (en) * | 2022-06-30 | 2024-01-04 | 西安交通大学 | Automatic seismic facies identification method based on combination of self-attention mechanism and u-shaped structure |
| CN115223009A (en) * | 2022-07-07 | 2022-10-21 | 云南电网有限责任公司电力科学研究院 | Small target detection method and device based on improved YOLOv5 |
| CN115471670A (en) * | 2022-07-20 | 2022-12-13 | 中国科学院西安光学精密机械研究所 | Space target detection method based on improved YOLOX network model |
| CN115376028A (en) * | 2022-08-24 | 2022-11-22 | 四川九洲空管科技有限责任公司 | Target detection method based on dense feature point splicing and improved YOLOV5 |
| CN115731533A (en) * | 2022-11-29 | 2023-03-03 | 淮阴工学院 | Vehicle-mounted target detection method based on improved YOLOv5 |
| CN115964653A (en) * | 2022-12-27 | 2023-04-14 | 泰山学院 | Underwater sound signal classification method, system, equipment and storage medium |
| CN115984698A (en) * | 2023-01-05 | 2023-04-18 | 华南农业大学 | Litchi fruit growing period identification method based on improved YOLOv5 |
| CN116129329A (en) * | 2023-03-02 | 2023-05-16 | 南京易视萃智科技有限公司 | Early-stage tiny flame detection method based on deep-space deconvolution and Kalman filtering |
| CN116645328A (en) * | 2023-04-25 | 2023-08-25 | 浙江理工大学 | Intelligent detection method for surface defects of high-precision bearing ring |
| CN116580324A (en) * | 2023-05-08 | 2023-08-11 | 长春理工大学 | Yolov 5-based unmanned aerial vehicle ground target detection method |
| CN117115686A (en) * | 2023-07-27 | 2023-11-24 | 中南民族大学 | Urban low-altitude small unmanned aerial vehicle detection method and system based on improved YOLOv7 |
Non-Patent Citations (7)
| Title |
|---|
| Faxiang Zhang , Shaodong Jiang, Xiaolei Zhang, Chang Wang, Jiasheng Ni, and Gangding Peng.Ultra-Thin Fiber Laser Hydrophone With Static Pressure Equalization and Improved Response.IEEE PHOTONICS TECHNOLOGY LETTERS.2019,第31卷(第24期),全文. * |
| Raja Sunkara and Tie Luo.No More Strided Convolutions or Pooling: A New CNN Building Block for Low-Resolution Images and Small Objects.European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases.2022,全文. * |
| Sanghyun Woo, Jongchan Parky, Joon-Young Lee, and In So Kweon.CBAM: Convolutional Block Attention Module.15th European Conference on Computer Vision.2018,全文. * |
| Yong Kang ZHENG，Jia Lin LIU.Bird Detection Algorithm for Transmission Line Based on Improved YOLOv7.2023 Panda Forum on Power and Energy.2023,全文. * |
| 姜劭栋，王蒙，赵庆超，张发祥，陈玉，王昌.一种用于油气井持气率测量的高可靠性光纤探针.半导体光电.2020,第41卷(第5期),全文. * |
| 张为 ; 魏晶晶 ; .嵌入DenseNet结构和空洞卷积模块的改进YOLO v3火灾检测算法.天津大学学报(自然科学与工程技术版).2020,(09),全文. * |
| 彭书博 ，陈兵旗 ，李景彬 ，范鹏宣 ，刘翔业，房鑫 ，邓红涛 ，张雄楚 .基于改进YOLOv7 的果园行间导航线检测.农 业 工 程 学 报.2023,第39卷(第16期),全文. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117668669A (en) | 2024-03-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN117668669B (en) | Pipeline safety monitoring method and system based on improvement YOLOv (YOLOv) | |
| CN110995339B (en) | Method for extracting and identifying time-space information of distributed optical fiber sensing signal | |
| CN107256635B (en) | Vehicle identification method based on distributed optical fiber sensing in intelligent traffic | |
| KR101895835B1 (en) | Ground penetrating radar survey system | |
| CN112069912B (en) | Method for identifying construction threat event of optical cable channel based on phi-OTDR | |
| CN116805061B (en) | Leakage event judging method based on optical fiber sensing | |
| CN112985574B (en) | High-precision classification identification method for optical fiber distributed acoustic sensing signals based on model fusion | |
| CN115731545A (en) | Cable tunnel inspection method and device based on fusion perception | |
| CN115060184A (en) | Optical fiber perimeter intrusion detection method and system based on recursive graph | |
| CN115359301A (en) | Data mining method based on cloud platform | |
| CN117726958B (en) | Intelligent detection and hidden danger identification method for inspection image target of unmanned aerial vehicle of distribution line | |
| CN111951505B (en) | Fence vibration intrusion positioning and mode identification method based on distributed optical fiber system | |
| CN116186642B (en) | Distributed optical fiber sensing event early warning method based on multidimensional feature fusion | |
| Wang et al. | Intelligent detection and recognition of multi-vibration events based on distributed acoustic sensor and improved YOLOv8 model | |
| CN113405646A (en) | Distributed vibration identification method based on dual-channel phi-OTDR (optical time Domain reflectometer) underground optical cable | |
| CN116504005A (en) | Perimeter security intrusion signal identification method based on improved CDIL-Bi-LSTM | |
| CN116935215A (en) | Road underground structural defect detection method, medium and system | |
| CN114354744B (en) | Power cable external damage prevention monitoring method, system and device | |
| Xiao et al. | Intrusion detection for high-speed railway system: a faster R-CNN approach | |
| Alsharafi et al. | Investigation on a Vision-Based Approch For Smart Pothole Detection Using Deep learning Based on Fast CNN | |
| CN114881085B (en) | Pattern recognition system based on space-time-frequency characteristic pyramid network | |
| CN112071009A (en) | Optical fiber pipeline early warning system and method thereof | |
| CN118172647B (en) | Fusion model debris flow infrasound recognition method based on deep learning | |
| CN117390413B (en) | Recognition method for distributed power optical fiber vibration signal noise reduction and time sequence feature extraction | |
| CN118277862A (en) | Distributed optical fiber sensing mode identification method and system based on neural network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |