CN116778329A - Urban road underground shallow disease detection method, device, equipment and medium - Google Patents
Urban road underground shallow disease detection method, device, equipment and medium Download PDFInfo
- Publication number
- CN116778329A CN116778329A CN202310744884.0A CN202310744884A CN116778329A CN 116778329 A CN116778329 A CN 116778329A CN 202310744884 A CN202310744884 A CN 202310744884A CN 116778329 A CN116778329 A CN 116778329A
- Authority
- CN
- China
- Prior art keywords
- disease
- radar
- training
- identification model
- features
- 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.)
- Pending
Links
- 201000010099 disease Diseases 0.000 title claims abstract description 152
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 title claims abstract description 152
- 238000001514 detection method Methods 0.000 title abstract description 22
- 238000012549 training Methods 0.000 claims abstract description 95
- 238000000034 method Methods 0.000 claims abstract description 77
- 238000013528 artificial neural network Methods 0.000 claims abstract description 17
- 238000002372 labelling Methods 0.000 claims abstract description 5
- 238000001228 spectrum Methods 0.000 claims description 32
- 230000004927 fusion Effects 0.000 claims description 20
- 238000000605 extraction Methods 0.000 claims description 12
- 230000007246 mechanism Effects 0.000 claims description 11
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 8
- 239000011800 void material Substances 0.000 claims description 6
- 238000004891 communication Methods 0.000 claims description 5
- 238000004590 computer program Methods 0.000 claims description 3
- 230000003902 lesion Effects 0.000 claims description 2
- 238000004458 analytical method Methods 0.000 abstract description 4
- 238000011156 evaluation Methods 0.000 abstract description 3
- 230000006870 function Effects 0.000 description 16
- 238000005259 measurement Methods 0.000 description 14
- 238000010586 diagram Methods 0.000 description 10
- 230000008569 process Effects 0.000 description 9
- 238000003062 neural network model Methods 0.000 description 7
- 230000000149 penetrating effect Effects 0.000 description 6
- 230000005540 biological transmission Effects 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 5
- 238000012545 processing Methods 0.000 description 5
- 230000002159 abnormal effect Effects 0.000 description 4
- 238000012986 modification Methods 0.000 description 4
- 230000004048 modification Effects 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000005070 sampling Methods 0.000 description 3
- 238000009412 basement excavation Methods 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000003628 erosive effect Effects 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 230000002829 reductive effect Effects 0.000 description 2
- 239000002689 soil Substances 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000003416 augmentation Effects 0.000 description 1
- 230000003190 augmentative effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000011797 cavity material Substances 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000007499 fusion processing Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 230000000670 limiting effect Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000004806 packaging method and process Methods 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 230000001902 propagating effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 230000002441 reversible effect Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
- G06V10/806—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The embodiment of the application provides a method, a device, equipment and a medium for detecting underground shallow diseases of an urban road, wherein the method comprises the following steps: according to the radar image acquired by the three-dimensional geological radar and the positioning data acquired by the positioning equipment, generating radar map data; labeling disease types of the radar map data, and generating a training sample set; and training to obtain a disease identification model by adopting the training sample set and a preset neural network algorithm. In this scheme, on the one hand, can realize discernment urban road underground shallow disease through disease recognition model, promote recognition efficiency, and combine radar map data to discern, also can ensure the degree of accuracy of discernment. On the other hand, the method is based on the cloud server, can timely and accurately carry out comprehensive analysis and evaluation on the road underground shallow disease, automatically pushes the detection result to related departments, and reduces labor cost and time cost.
Description
Technical Field
The application relates to the technical field of underground engineering detection, in particular to a method, a device, equipment and a medium for detecting underground shallow diseases of urban roads.
Background
In recent years, with the increasing development and utilization intensity of urban underground space, phenomena such as underground water and soil loss and erosion of roads are caused by the factors such as repeated road excavation, damage and leakage of underground pipe networks, road surface load change, foundation pit construction disturbance and the like, so that diseases such as cavities are formed, and finally, road collapse accidents are frequently caused. If the hidden danger detection of the urban road collapse can be developed in advance, the occurrence of holes, void, rich water and the like in the underground shallow layer of the urban road can be discovered and prevented early, the loss of life and property can be reduced, and meanwhile, the urban road management level is improved.
In the prior art, nondestructive detection technologies such as a ground penetrating radar method, a multi-channel surface wave method, a high-density resistivity method, a seismic mapping method and the like are generally adopted for detecting the road underground shallow diseases. The ground penetrating radar is used as one of the most common technologies for engineering investigation and detection, is widely applied to engineering by the technical advantages of high resolution, high efficiency, wide detection range, no damage, low cost and the like, and has become the first choice technical method for detecting and identifying the underground shallow diseases of urban roads.
However, the interpretation work in the current ground penetrating radar method is mainly processed and interpreted by engineering experience of technicians, however, due to the large interpretation difficulty and high manual interpretation experience requirement, the method has the defects of more misjudgment, low speed, low precision, low efficiency and the like, thereby influencing the accuracy of detecting the underground shallow diseases of urban roads and the detection working efficiency.
Disclosure of Invention
The application aims to provide a training method, device, equipment and medium for an identification model of an underground shallow disease, aiming at the defects in the prior art, so as to solve the problem of low accuracy of underground shallow disease detection in the prior art.
In order to achieve the above purpose, the technical scheme adopted by the embodiment of the application is as follows:
in a first aspect, an embodiment of the present application provides a training method for an identification model of an underground shallow disease, where the method includes:
according to radar images acquired by the three-dimensional geological radar and positioning data acquired by the positioning equipment, radar spectrum data are generated, wherein the radar spectrum data comprise: the association relation between the radar image and the positioning data;
labeling disease types of the radar map data, and generating a training sample set;
and training to obtain a disease identification model by adopting the training sample set and a preset neural network algorithm, wherein the disease identification model is used for identifying the type of the underground shallow disease.
In an alternative embodiment, before the generating radar map data according to the radar image acquired by the three-dimensional geological radar and the positioning data acquired by the positioning device, the method further includes:
and respectively receiving the radar image and the positioning data transmitted by the three-dimensional geological radar and the positioning equipment through a 5G network.
In an optional implementation manner, the training to obtain the disease identification model by using the training sample set and a preset neural network algorithm includes:
extracting features of training samples in the training sample set by adopting a lightweight feature extraction network;
fusing the features by adopting a feature fusion algorithm to obtain fused features;
training based on a preset loss function and the fused features until the loss function converges, and obtaining the disease identification model, wherein an output layer of the disease identification model comprises a plurality of pre-measuring heads.
In an alternative embodiment, the feature fusion algorithm is used to fuse the features, and the obtaining the fused features includes:
adopting an attention mechanism to aggregate the features along two space directions respectively, and reserving the features to be fused;
and fusing the features to be fused by adopting a self-adaptive spatial feature fusion mechanism, and obtaining the fused features.
In an alternative embodiment, the method further comprises:
installing a preset engine and deploying a preset frame to build an environment for deploying the disease identification model.
In an optional implementation manner, after training to obtain the disease identification model by using the training sample set and a preset neural network algorithm, the method further includes:
collecting radar map data of a target area;
and inputting the radar spectrum data into the disease identification model, and obtaining the type of the underground shallow disease corresponding to the radar spectrum data.
In an alternative embodiment, the types of subsurface diseases include: hollow, void, water-rich, generally loose, severely loose.
In a second aspect, another embodiment of the present application provides an apparatus for training an identification model of a disease in a shallow subsurface, the apparatus comprising:
the acquisition module is used for acquiring radar images according to the three-dimensional geological radar and positioning data acquired by the positioning equipment;
the generation module is used for marking the disease types of the radar map data and generating a training sample set;
the identification module is used for training and obtaining a disease identification model by adopting the training sample set and a preset neural network algorithm, and the disease identification model is used for identifying the type of the underground shallow disease.
In an alternative embodiment, the acquisition module is further configured to receive the radar image and the positioning data transmitted by the three-dimensional geological radar and the positioning device through the 5G network, respectively.
In an alternative embodiment, the identification module is configured to extract features of the training samples in the training sample set using a lightweight feature extraction network;
fusing the features by adopting a feature fusion algorithm to obtain fused features;
training based on a preset loss function and the fused features until the loss function converges, and obtaining the disease identification model, wherein an output layer of the disease identification model comprises a plurality of pre-measuring heads.
In an alternative embodiment, the identifying module is configured to aggregate the features along two spatial directions respectively by using an attention mechanism, and reserve features to be fused;
and fusing the features to be fused by adopting a self-adaptive spatial feature fusion mechanism, and obtaining the fused features.
In an alternative embodiment, the deployment module is configured to install a preset engine and deploy a preset frame to build an environment for deploying the disease identification model.
In an alternative embodiment, the identification module is further configured to collect radar spectrum data of the target area;
and inputting the radar spectrum data into the disease identification model, and obtaining the type of the underground shallow disease corresponding to the radar spectrum data.
In an alternative embodiment, the types of subsurface diseases include: hollow, void, water-rich, generally loose, severely loose.
In a third aspect, another embodiment of the present application provides an electronic device, including: a processor, a storage medium and a bus, the storage medium storing machine-readable instructions executable by the processor, the processor in communication with the storage medium via the bus when the electronic device is operating, the processor executing the machine-readable instructions to perform the steps of the method of model training for identification of subsurface shallow lesions as described in any of the first aspect above.
In a fourth aspect, another embodiment of the present application provides a storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of the method for training a model for identification of a shallow subsurface disease as described in any one of the first aspects above.
The beneficial effects of the application are as follows:
the application provides a training method, a training device, training equipment and training media for an identification model of an underground shallow disease, wherein the training method comprises the following steps: according to the radar image acquired by the three-dimensional geological radar and the positioning data acquired by the positioning equipment, generating radar map data; labeling disease types of radar map data, and generating a training sample set; and training to obtain a disease identification model by adopting a training sample set and a preset neural network algorithm. In this scheme, on the one hand, can realize discernment urban road underground shallow disease through disease recognition model, promote recognition efficiency, and combine radar map data to discern, also can ensure the degree of accuracy of discernment. On the other hand, the method is based on the cloud server, can timely and accurately carry out comprehensive analysis and evaluation on the road underground shallow disease, automatically pushes the detection result to related departments, and reduces labor cost and time cost.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the embodiments will be briefly described below, it being understood that the following drawings only illustrate some embodiments of the present application and therefore should not be considered as limiting the scope, and other related drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic flow chart of a training method for an identification model of an underground shallow disease according to an embodiment of the present application;
FIG. 2 is a schematic flow chart of another method for training an identification model of a shallow disease in the subsurface provided by the embodiment of the application;
fig. 3 is a schematic diagram of a network structure of a disease recognition model according to an embodiment of the present application;
FIG. 4 is a flow chart of another method for training an identification model of a shallow disease in the subsurface provided by the embodiment of the application;
FIG. 5 is a schematic flow chart of another method for training an identification model of a shallow disease in the subsurface provided by the embodiment of the application;
FIG. 6 is a schematic diagram of an identification system for shallow subsurface diseases according to an embodiment of the present application;
fig. 7 is a schematic structural diagram of an apparatus for training an identification model of an underground shallow disease according to an embodiment of the present application;
fig. 8 is a schematic structural diagram of an electronic device according to an embodiment of the present application.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present application more apparent, the technical solutions of the embodiments of the present application will be clearly and completely described with reference to the accompanying drawings in the embodiments of the present application, and it should be understood that the drawings in the present application are for the purpose of illustration and description only and are not intended to limit the scope of the present application. In addition, it should be understood that the schematic drawings are not drawn to scale. A flowchart, as used in this disclosure, illustrates operations implemented according to some embodiments of the present application. It should be understood that the operations of the flow diagrams may be implemented out of order and that steps without logical context may be performed in reverse order or concurrently. Moreover, one or more other operations may be added to or removed from the flow diagrams by those skilled in the art under the direction of the present disclosure.
In addition, the described embodiments are only some, but not all, embodiments of the application. The components of the embodiments of the present application generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the application, as presented in the figures, is not intended to limit the scope of the application, as claimed, but is merely representative of selected embodiments of the application. All other embodiments, which can be made by a person skilled in the art without making any inventive effort, are intended to be within the scope of the present application.
It should be noted that the term "comprising" will be used in embodiments of the application to indicate the presence of the features stated hereafter, but not to exclude the addition of other features.
In recent years, with the increasing development and utilization intensity of urban underground space, phenomena such as underground water and soil loss and erosion of roads are caused by the factors such as repeated road excavation, damage and leakage of underground pipe networks, road surface load change, foundation pit construction disturbance and the like, so that diseases such as cavities are formed, and finally, road collapse accidents are frequently caused. Therefore, the method for detecting the hidden danger of the urban road collapse is usually developed in advance to avoid the occurrence of road accidents. In the prior art, a ground penetrating radar method is generally adopted, but the ground penetrating radar method is often dependent on engineering experience of technicians to process and interpret, and has the problems of low accuracy of obtained results and low working efficiency.
In view of the above, the embodiment of the application provides a method for detecting urban road underground shallow diseases, which is based on a cloud server, and can effectively improve the accuracy of detection results by combining a geological radar detection technology and an artificial intelligent image recognition technology.
Fig. 1 is a flow chart of a training method for an identification model of an underground shallow disease according to an embodiment of the present application, where an execution subject of the method is a cloud server, as shown in fig. 1, and the method may include:
s101, generating radar map data according to radar images acquired by the three-dimensional geological radar and positioning data acquired by the positioning equipment.
The working principle of the three-dimensional geological radar is that high-frequency pulse electromagnetic waves are emitted to an underground medium, and the high-frequency pulse electromagnetic waves are used for determining the distribution condition of the underground medium. Specifically, based on the electrical difference of the underground medium, the three-dimensional geological radar can emit high-frequency pulse electromagnetic waves through one antenna, and the other antenna receives electromagnetic wave signals reflected by the underground medium and analyzes and interprets the received electromagnetic wave signals to obtain radar images.
In the specific implementation, electromagnetic waves propagating in an underground medium encounter media with different dielectric constants or are reflected by a measured object body to form different electromagnetic wave propagation paths and waveform information, so that the three-dimensional geological radar can analyze the spatial position of the underground object body or interface according to the information such as the propagation time, waveform, amplitude and the like in a received electromagnetic wave signal, and further a corresponding radar image can be obtained.
Optionally, the radar image is a cross-sectional view of the subsurface shallow layer obtained through processing analysis. If there is a disease or an abnormal body in the underground medium, the difference in dielectric constant between the two sides of the interface of the disease or abnormal body is large, and a strong reflection signal is generated, so that the position of the disease or abnormal body can be determined from the time section of the section.
On the other hand, the positioning device can be used for precisely determining the three-dimensional coordinates of the three-dimensional geological radar antenna, and meanwhile, acquiring the measurement time of the corresponding measuring point to serve as corresponding positioning data.
When the radar image is acquired, the geographic position of the road is acquired through the positioning equipment, so that the mapping relation of the road position is formed, and the association relation between the radar image and the positioning data can be determined. If diseases or abnormal bodies exist in the underground medium, corresponding positioning data can be searched according to the association relation.
It should be noted that, the embodiment of the present application may also collect data according to two-dimensional geological radar, which is not limited herein. In addition, the positioning device according to the embodiment of the present application may be based on a global positioning system (Global Positioning System, abbreviated as GPS) positioning technology, or may be a beidou satellite navigation system in china, a galileo satellite navigation system in the european union, a russian global navigation satellite system, or the like, which is not limited herein.
S102, marking disease types of radar map data, and generating a training sample set.
In the training process, the disease type of the radar map data is known, for example, the disease type of some positions can be known through human or historical data. Optionally, after the cloud server receives the original data of the radar spectrum data, the cloud server can obtain the radar spectrum data in a standard format which is available by an artificial intelligent algorithm through processing, so that the disease type is marked, and a road disease radar spectrum data sample base is constructed, namely a training sample set is obtained.
Optionally, a visualized image calibration tool (e.g., labelImg) is used for marking radar spectrum data, the position of the underground shallow disease is marked in a box form, the disease type is marked, and the radar spectrum data in the Pascal VOC format is prepared, so that a training sample set is generated. The marked disease types can comprise non-disease, cavity, void, water-rich, general loose, serious loose and the like.
Optionally, the training sample set is subjected to an augmentation operation, and the augmented training sample set is divided into a training set, a verification set and a test set according to a certain proportion, wherein the specific proportion value is set according to an actual application scene, and the method is not limited. The method can further expand the database in subsequent application, enrich the training sample set and improve the accuracy of the intelligent recognition algorithm.
S103, training to obtain a disease identification model by adopting a training sample set and a preset neural network algorithm.
Wherein, the disease recognition model is used for recognizing the type of the underground shallow disease.
Optionally, the preset neural network algorithm is optimized and trained by adopting an improved YOLOv5 neural network model. In order to better understand the application, firstly, a basic network architecture of the existing YOLOv5 neural network model is described, the YOLOv5 network mainly comprises four parts, namely Input, backbone, neck and Head, firstly, a picture is preprocessed through a series of operations such as data enhancement of the Input part and then is sent into the Backbone, and the Backbone extracts characteristics of the processed picture; then, the extracted features are subjected to Neck module feature fusion processing to obtain features with three sizes of large, medium and small; finally, the fused features are sent to a detection head, and the result is obtained after detection.
It should be noted that YOLO is a classical single-stage object detection algorithm, which converts the object detection problem into a regression problem of middle boundary separation in space, and the processing speed of YOLO based on object detection is very fast, and the recognition degree of background and object is very high. YOLOv5 is the latest series of YOLOv, and the embodiment of the application takes the model structure of YOLOv5 as a main reference, on one hand, because the model structure of YOLOv5 has better results, the accuracy of the overall results is more ensured; on the other hand, the model is lighter, meets the requirements of practical application and training research of the embodiment of the application, and is convenient for the deployment of the model on a cloud server and the implementation of detection tasks.
Optionally, the training sample set is brought into a preset neural network algorithm to carry out parameter adjustment, so that the disease identification model is obtained through training.
After the disease recognition model is obtained, the disease recognition model can be tested and verified by the test set and the verification set, and if the disease recognition model is unqualified, the disease recognition model can be further adjusted by training until the disease recognition model is qualified.
In the embodiment, radar map data are generated according to radar images acquired by the three-dimensional geological radar and positioning data acquired by the positioning equipment; labeling disease types of radar map data, and generating a training sample set; the disease recognition model is obtained through training by adopting the training sample set and the preset neural network algorithm, so that the urban road underground shallow disease can be recognized through the disease recognition model, the recognition efficiency is improved, the radar map data is combined for recognition, and the recognition accuracy can be ensured. On the other hand, the method is based on the cloud server, can timely and accurately carry out comprehensive analysis and evaluation on the road underground shallow disease, automatically pushes the detection result to related departments, and reduces labor cost and time cost.
Optionally, in the step S101, before generating radar map data according to the radar image acquired by the three-dimensional geological radar and the positioning data acquired by the positioning device, the method further includes: and respectively receiving radar images and positioning data transmitted by the three-dimensional geological radar and the positioning equipment through the 5G network.
In this embodiment, an internet of things device is installed on a three-dimensional geological radar device, and the internet of things device is used for packaging acquired radar images and positioning data, sending the radar images and the positioning data to a 5G mobile base station through an internet of things card, uploading the radar images and the positioning data to a gateway, distributing the data through cloud computing, and sending the data to a cloud server. The method solves the problems that in the traditional method, because the data volume acquired by the ground penetrating radar operation is large, the equipment data transmission is disconnected at any time when transmitted through the common network, and the data uploading interval time is long.
Fig. 2 is a flow chart of another method for training an identification model of a shallow disease in the underground, and fig. 3 is a network structure diagram of an identification model of a disease, which is provided in the embodiment of the application. As shown in fig. 2, in step 103, a disease identification model is obtained by training using a training sample set and a preset neural network algorithm, and the method includes:
s201, extracting features of training samples in a training sample set by adopting a lightweight feature extraction network.
Optionally, substituting the training sample set into the improved YOLOv5 neural network model to perform feature extraction. Referring to fig. 3, the disease identification model provided in the embodiment of the present application may include: the embodiment of the application replaces the feature extraction network in the original YOLOv5 feature extraction layer with a lightweight feature extraction network (for example, a MobileNetv3 lightweight feature extraction skeleton network 101) so that the environment perception is more accurate and the memory occupied by a model is smaller, wherein the feature extraction layer comprises a CB-h module and a V3-Res module, and the CB-h consists of Conv+BH+h-swish layers.
The V3-Res modules in the MobileNetv3 lightweight feature extraction skeleton network are composed of structural units using a residual network, and the number of V3-Res modules can be set according to the network depth, as shown in fig. 3, and 15V 3-Res modules are set in total in the embodiment of the application.
S202, fusing the features by adopting a feature fusion algorithm to obtain fused features.
And S203, training based on the preset loss function and the fused features until the loss function converges, and obtaining a disease identification model.
The output layer of the disease recognition model comprises a plurality of pre-measuring heads. It should be noted that, in order to alleviate the negative effect caused by the target scale change after the feature fusion, a group of smaller fourth pre-measurement heads P2 are added on the basis of the original YOLOv5 neural network model in the embodiment of the present application. Referring to fig. 3, the improved YOLOv5 neural network model in the embodiment of the present application retains the original first pre-measurement head P5, second pre-measurement head P4, third pre-measurement head P3, and newly introduces the fourth pre-measurement head P2, where the first pre-measurement head P5, second pre-measurement head P4, third pre-measurement head P3 are respectively used for identifying the shallow disease areas in the large area, the middle area, and the small area, and the introduced fourth pre-measurement head P2 is mainly used for identifying the shallow disease areas in the small area, that is, the embodiment of the present application can implement identification of the shallow disease areas in the multi-size underground, so as to improve the applicability of the method of the present application. It should be noted that the specific sizes of the large area, the medium area, the small area and the micro area are not limited, and the present application can be flexibly set according to actual application scenarios.
Optionally, the foregoing preset loss function may select an EIoU loss function, and calculate the distribution difference of the forward modeling data and the real disease data by using the EIoU loss function instead of the CIoU loss function, so as to effectively measure the distances between the calculated forward modeling data and the real disease data in the feature space.
Fig. 4 is a flow chart of another training method for an identification model of a shallow disease in the underground, as shown in fig. 4, in the step S202, the feature is fused by using a feature fusion algorithm, so as to obtain a fused feature, which includes:
s301, adopting an attention mechanism to aggregate the features along two space directions respectively, and reserving the features to be fused.
Optionally, an attention mechanism (Coordinate Attention, abbreviated as CA) is embedded in the YOLOv5 neural network model, and the channel attention can be decomposed into two one-dimensional feature encoding processes, which respectively aggregate features along two spatial directions, wherein remote dependencies can be captured along one spatial direction, while precise location information can be retained along the other spatial direction. This approach can enhance sensitivity to information such as direction and position.
S302, adopting a self-adaptive spatial feature fusion mechanism to fuse features to be fused and obtaining the fused features.
The adaptive spatial feature fusion (Adaptively Spatial Feature Fusion, abbreviated as ASFF) is a fusion mode capable of adaptively learning spatial weights of the map fusion of the features of different scales and fully utilizing the features of different scales. According to the embodiment of the application, the Mosaic image enhancement strategy is improved in the original YOLOv5 network structure, and ASFF is added in the Neck connection part of the network structure, so that the characteristics of different scales can be fully utilized, and the detection performance of a small target can be enhanced.
With continued reference to fig. 3, the disease identification model provided in the embodiment of the present application introduces a CA module and an ASFF module based on the existing neg module, where each of the neg modules may be provided with a CA module after fusing the feature function units Concat, as shown in fig. 3, and includes 7 CA modules; further, from the 4 th CA module (corresponding to the number 27), each CA module may correspond to an ASFF module, and outputs of the 4 th, 5 th, 6 th and 7 th CA modules (corresponding to the numbers 27, 30, 33 and 36, respectively) may be given to each ASFF module, and each ASFF module may correspond to one convolution unit Conv, as shown in fig. 3, including 4 convolution units, wherein the outputs of the four convolution units are input to the first pre-measurement head P5, the second pre-measurement head P4, the third pre-measurement head P3 and the fourth pre-measurement head P2, respectively. Furthermore, it should be noted that CBS in the neg block represents convolution units, which may be based on convolution functions, BN, siLU, and; upsamples represent upsampling units, and it can be seen from the figure that the output of the convolution unit CBS can be used as the input of the upsampling unit upsamples, and the output of the upsampling unit upsamples can be used as the input of the fusion feature function unit Concat.
In the embodiment, the characteristics can be fused by adopting self-adaptive spatial characteristic fusion, and the method enhances the network characteristic characterization capability.
Optionally, the method of the embodiment of the application can realize model training at the cloud server, and further can realize model calling at the cloud server, namely, the type of the underground shallow disease is identified through the trained disease identification model. Then on the cloud service, the method further comprises installing a preset engine and deploying a preset frame to build an environment for deploying the disease identification model.
Optionally, a preset engine (for example, dock) may be installed on the cloud service platform, the improved YOLOv5 neural network model and the labeled radar spectrum data constructed in step S102 are deployed to the cloud server, and the training process is completed by using the GPU of the cloud server.
Furthermore, an algorithm environment can be built on a cloud server by using a proper deployment framework such as Tensorrt, onnx-run time and ncnn, and finally model deployment is carried out, and a road disease pattern recognition algorithm model file is copied to a container to be deployed on a cloud platform, so that automatic recognition of a road disease radar pattern is realized.
Fig. 5 is a flow chart of another method for training an identification model of a shallow disease in the underground, as shown in fig. 5, in the step S103, after training to obtain the identification model of the disease by using the training sample set and a preset neural network algorithm, the method further includes:
s401, acquiring radar spectrum data of a target area.
Optionally, after the disease recognition model is trained, radar spectrum data of a target area can be recognized, wherein the target area is a road section to be detected in the urban road.
It should be noted that, the radar spectrum data of the target area may be obtained with reference to the embodiment shown in fig. 1, that is, the radar spectrum data is obtained according to the radar image collected by the three-dimensional geological radar and the positioning data collected by the positioning device, which is not described herein again.
S402, inputting radar spectrum data into a disease identification model, and obtaining the type of the underground shallow disease corresponding to the radar spectrum data.
Optionally, the disease identification model may output probabilities of different types of underground shallow disease, and select a type with the highest probability as a type of underground shallow disease corresponding to the radar map data.
Further, after the type of the underground shallow disease corresponding to the acquired radar map data is identified, specific disease information can be marked and transmitted to terminal equipment of a specific management department. The disease information may include: the type of the underground shallow disease, the disease position coordinates, the disease level, the sampling time, the sampling unit, the sampling personnel and the like corresponding to the radar map data. The terminal device of the management department can be, for example, a terminal device of a road maintenance department, and can receive the disease information on the client or software through the related account number for timely processing.
Fig. 6 is a schematic diagram of an underground shallow disease recognition system provided by the embodiment of the present application, where the system may implement training, calling, information transmission, etc. of the above-mentioned underground shallow disease recognition model, and as shown in fig. 6, the system includes: data acquisition unit 501, data transmission unit 502, cloud server 503:
the data acquisition unit 501 may be in communication connection with a three-dimensional geological radar and a positioning device, and is configured to acquire a radar image acquired by the three-dimensional geological radar and positioning data acquired by the positioning device.
The data transmission unit 502 is located between the data acquisition unit 501 and the cloud server 503, and can build a 5G transmission path for transmitting data to the cloud server 503.
The cloud server 503 is configured to perform the above method, and may transmit disease information to the terminal device.
The cloud server 503 may have a receiving unit that is responsible for receiving radar spectrum data, a unit that processes and normalizes the data, and a unit that is also provided with a training model and a usage model, which are not described herein.
It will be appreciated that the identification system of subsurface shallow disease may be mounted on a daily inspection vehicle by collecting data at different times and/or locations and identifying the type of subsurface shallow disease.
In the embodiment, the system is based on the cloud server, and can be used for comprehensively analyzing and evaluating the road underground shallow diseases timely and accurately and automatically pushing the detection result to related departments, so that the labor cost and the time cost are reduced.
Fig. 7 is a schematic structural diagram of an apparatus for training an identification model of an underground shallow disease according to an embodiment of the present application, as shown in fig. 7, the apparatus 600 for training an identification model may include:
the acquisition module 601 is configured to generate radar spectrum data according to a radar image acquired by the three-dimensional geological radar and positioning data acquired by the positioning device, where the radar spectrum data includes: the association relation between the radar image and the positioning data;
the generating module 602 is configured to label the disease type of the radar spectrum data, and generate a training sample set;
the identifying module 603 is configured to train and obtain a disease identifying model by using the training sample set and a preset neural network algorithm, where the disease identifying model is used to identify a type of a shallow underground disease.
In an alternative embodiment, the acquisition module 601 is further configured to receive the radar image and the positioning data transmitted by the three-dimensional geological radar and the positioning device through the 5G network, respectively.
In an alternative embodiment, the identifying module 603 is configured to extract features of the training samples in the training sample set using a lightweight feature extraction network;
fusing the features by adopting a feature fusion algorithm to obtain fused features;
training based on a preset loss function and the fused features until the loss function converges, and obtaining the disease identification model, wherein an output layer of the disease identification model comprises a plurality of pre-measuring heads.
In an alternative embodiment, the identifying module 603 is configured to aggregate the features along two spatial directions respectively by using an attention mechanism, and reserve features to be fused;
and fusing the features to be fused by adopting a self-adaptive spatial feature fusion mechanism, and obtaining the fused features.
In an alternative embodiment, the deployment module is configured to install a preset engine and deploy a preset frame to build an environment for deploying the disease identification model.
In an alternative embodiment, the identifying module 603 is further configured to collect radar spectrum data of the target area;
and inputting the radar spectrum data into the disease identification model, and obtaining the type of the underground shallow disease corresponding to the radar spectrum data.
In an alternative embodiment, the types of subsurface diseases include: hollow, void, water-rich, generally loose, severely loose.
The process flow of each module in the apparatus and the interaction flow between the modules may be described with reference to the related descriptions in the above method embodiments, which are not described in detail herein.
Fig. 8 is a schematic structural diagram of an electronic device according to an embodiment of the present application, where the electronic device may be integrated in the recognition model training apparatus, as shown in fig. 8, and the electronic device may include: a processor 701, a storage medium 702, and a bus 703, the storage medium 702 storing machine-readable instructions executable by the processor 701, the processor 701 and the storage medium 702 communicating over the bus 703 when the electronic device is running, the processor 701 executing the machine-readable instructions to perform the steps of the method embodiments described above. The specific implementation manner and the technical effect are similar, and are not repeated here.
Optionally, the present application further provides a storage medium, on which a computer program is stored, which when being executed by a processor performs the steps of the above-described method embodiments. The specific implementation manner and the technical effect are similar, and are not repeated here.
In the several embodiments provided by the present application, it should be understood that the disclosed apparatus and method may be implemented in other manners. For example, the apparatus embodiments described above are merely illustrative, e.g., the division of the units is merely a logical function division, and there may be additional divisions when actually implemented, e.g., multiple units or components may be combined or integrated into another system, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other form.
The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional unit in the embodiments of the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in hardware plus software functional units.
The integrated units implemented in the form of software functional units described above may be stored in a computer readable storage medium. The software functional unit is stored in a storage medium, and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) or a processor (english: processor) to perform some of the steps of the methods according to the embodiments of the application. And the aforementioned storage medium includes: u disk, mobile hard disk, read-Only Memory (ROM), random access Memory (Random Access Memory, RAM), magnetic disk or optical disk, etc.
It should be noted that in this document, relational terms such as "first" and "second" and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
The above description is only of the preferred embodiments of the present application and is not intended to limit the present application, but various modifications and variations can be made to the present application by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application. It should be noted that: like reference numerals and letters denote like items in the following figures, and thus once an item is defined in one figure, no further definition or explanation thereof is necessary in the following figures. The above description is only of the preferred embodiments of the present application and is not intended to limit the present application, but various modifications and variations can be made to the present application by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application.
Claims (10)
1. An identification model training method for an underground shallow disease is characterized by being applied to a cloud server, and comprises the following steps:
according to radar images acquired by the three-dimensional geological radar and positioning data acquired by the positioning equipment, radar spectrum data are generated, wherein the radar spectrum data comprise: the association relation between the radar image and the positioning data;
labeling disease types of the radar map data, and generating a training sample set;
and training to obtain a disease identification model by adopting the training sample set and a preset neural network algorithm, wherein the disease identification model is used for identifying the type of the underground shallow disease.
2. The method of claim 1, wherein prior to generating radar map data from the radar image acquired by the three-dimensional geological radar and the positioning data acquired by the positioning device, the method further comprises:
and respectively receiving the radar image and the positioning data transmitted by the three-dimensional geological radar and the positioning equipment through a 5G network.
3. The method according to claim 1, wherein training to obtain the disease identification model using the training sample set and a predetermined neural network algorithm comprises:
extracting features of training samples in the training sample set by adopting a lightweight feature extraction network;
fusing the features by adopting a feature fusion algorithm to obtain fused features;
training based on a preset loss function and the fused features until the loss function converges, and obtaining the disease identification model, wherein an output layer of the disease identification model comprises a plurality of pre-measuring heads.
4. A method according to claim 3, wherein the fusing the features using a feature fusion algorithm to obtain fused features comprises:
adopting an attention mechanism to aggregate the features along two space directions respectively, and reserving the features to be fused;
and fusing the features to be fused by adopting a self-adaptive spatial feature fusion mechanism, and obtaining the fused features.
5. The method according to claim 1, wherein the method further comprises:
installing a preset engine and deploying a preset frame to build an environment for deploying the disease identification model.
6. The method according to any one of claims 1-5, wherein after training to obtain the disease identification model using the training sample set and a preset neural network algorithm, further comprising:
collecting radar map data of a target area;
and inputting the radar spectrum data into the disease identification model, and obtaining the type of the underground shallow disease corresponding to the radar spectrum data.
7. The method of claim 6, wherein the type of subsurface disease comprises: hollow, void, water-rich, generally loose, severely loose.
8. An identification model training device for an underground shallow disease, which is characterized by comprising:
the acquisition module is used for generating radar spectrum data according to radar images acquired by the three-dimensional geological radar and positioning data acquired by the positioning equipment, and the radar spectrum data comprises: the association relation between the radar image and the positioning data;
the generation module is used for marking the disease types of the radar map data and generating a training sample set;
the identification module is used for training and obtaining a disease identification model by adopting the training sample set and a preset neural network algorithm, and the disease identification model is used for identifying the type of the underground shallow disease.
9. An electronic device, comprising: a processor, a storage medium and a bus, said storage medium storing machine-readable instructions executable by said processor, said processor in communication with said storage medium via the bus when the electronic device is running, said processor executing said machine-readable instructions to perform the steps of the method of model training for identification of subsurface shallow lesions as in any of claims 1-7.
10. A computer readable storage medium, characterized in that the computer readable storage medium has stored thereon a computer program which, when executed by a processor, implements the steps of the method for training a model for identifying a shallow subsurface disease as claimed in any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310744884.0A CN116778329A (en) | 2023-06-21 | 2023-06-21 | Urban road underground shallow disease detection method, device, equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310744884.0A CN116778329A (en) | 2023-06-21 | 2023-06-21 | Urban road underground shallow disease detection method, device, equipment and medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116778329A true CN116778329A (en) | 2023-09-19 |
Family
ID=87985487
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310744884.0A Pending CN116778329A (en) | 2023-06-21 | 2023-06-21 | Urban road underground shallow disease detection method, device, equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116778329A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117233752A (en) * | 2023-11-08 | 2023-12-15 | 江苏筑升土木工程科技有限公司 | Road underground disease body water content calculation and analysis method based on radar nondestructive detection |
| CN117690093A (en) * | 2024-01-31 | 2024-03-12 | 华能澜沧江水电股份有限公司 | Dam safety monitoring operation maintenance method and device, electronic equipment and storage medium |
-
2023
- 2023-06-21 CN CN202310744884.0A patent/CN116778329A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117233752A (en) * | 2023-11-08 | 2023-12-15 | 江苏筑升土木工程科技有限公司 | Road underground disease body water content calculation and analysis method based on radar nondestructive detection |
| CN117233752B (en) * | 2023-11-08 | 2024-01-30 | 江苏筑升土木工程科技有限公司 | Road underground disease body water content calculation and analysis method based on radar nondestructive detection |
| CN117690093A (en) * | 2024-01-31 | 2024-03-12 | 华能澜沧江水电股份有限公司 | Dam safety monitoring operation maintenance method and device, electronic equipment and storage medium |
| CN117690093B (en) * | 2024-01-31 | 2024-04-26 | 华能澜沧江水电股份有限公司 | Dam safety monitoring operation maintenance method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113256562B (en) | Road underground hidden danger detection method and system based on radar images and artificial intelligence | |
| CN116778329A (en) | Urban road underground shallow disease detection method, device, equipment and medium | |
| CN117272191B (en) | Urban underground space intelligent mapping method and system | |
| CA2395848A1 (en) | Radar cross-section measurement system for analysis of insulative structures | |
| EP2848962A1 (en) | Method for identifying and classifying petroleum exuding areas at the bottom of the sea by reverse modelling | |
| CN113360587B (en) | Land surveying and mapping equipment and method based on GIS technology | |
| Chen et al. | A novel image-based approach for interactive characterization of rock fracture spacing in a tunnel face | |
| Van Son et al. | A framework for reliable three-dimensional underground utility mapping for urban planning | |
| CN115423968B (en) | Power transmission channel optimization method based on point cloud data and live-action three-dimensional model | |
| Song et al. | Development of comprehensive accuracy assessment indexes for building footprint extraction | |
| US11754704B2 (en) | Synthetic-aperture-radar image processing device and image processing method | |
| CN117724089B (en) | Ground and underground integrated intelligent mobile detection system | |
| CN117437559A (en) | Unmanned aerial vehicle-based method and device for detecting ground surface rock movement deformation of coal mining area | |
| CN113433568B (en) | Laser radar observation simulation method and device | |
| Sefercik et al. | Area-based quality control of airborne laser scanning 3D models for different land classes using terrestrial laser scanning: sample survey in Houston, USA | |
| Comert et al. | Object based building extraction and building period estimation from unmanned aerial vehicle data | |
| CN115859081A (en) | Visual detection method and device for pipeline | |
| CN104077592B (en) | Automatic extraction method for high-resolution remote-sensing image navigation mark | |
| Marsella et al. | Terrestrial laser scanning survey in support of unstable slopes analysis: the case of Vulcano Island (Italy) | |
| CN115630492A (en) | Intelligent inversion method, system and storage medium for tunnel lining disease change characteristics | |
| CN114004740B (en) | Building wall line extraction method based on unmanned aerial vehicle laser radar point cloud | |
| CN112964938B (en) | Lightning single-station positioning method, device and system based on artificial intelligence | |
| Braga-Neto et al. | Spatial data for fungal specimens: retrospective georeferencing and practical recommendations for mycologists | |
| CA2403462A1 (en) | Method and system for identification of subterranean objects | |
| KR102677509B1 (en) | System for managing underground space integrated map capable of acquiring 3D information on underground facilities and management method thereof |
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 |