CN113284144B - Tunnel detection method and device based on unmanned aerial vehicle - Google Patents
Tunnel detection method and device based on unmanned aerial vehicle Download PDFInfo
- Publication number
- CN113284144B CN113284144B CN202110830687.1A CN202110830687A CN113284144B CN 113284144 B CN113284144 B CN 113284144B CN 202110830687 A CN202110830687 A CN 202110830687A CN 113284144 B CN113284144 B CN 113284144B
- Authority
- CN
- China
- Prior art keywords
- aerial vehicle
- unmanned aerial
- tunnel
- neural network
- network model
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- 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
- G06N3/045—Combinations of networks
-
- 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
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
Abstract
The invention discloses a tunnel detection method and a tunnel detection device based on an unmanned aerial vehicle, wherein the method comprises the following steps: receiving a video image of a tunnel acquired by an unmanned aerial vehicle; preprocessing and standardizing the video image; and inputting the video image after preprocessing and standardization processing into a trained three-dimensional convolutional neural network (3D-CNN) model, and outputting a tunnel disease detection result by the three-dimensional convolutional neural network model according to the video image. According to the invention, the video image of the tunnel is acquired by the unmanned aerial vehicle, and the video image is input into the trained three-dimensional convolutional neural network model, so that the characteristics of space and time in the video can be better captured, the tunnel disease detection result is rapidly output, the tunnel disease identification rate, the detection frequency and the processing speed are improved, and the artificial dependence degree is reduced.
Description
Technical Field
The invention relates to the technical field of tunnel detection, in particular to a tunnel detection method and device based on an unmanned aerial vehicle.
Background
With the acceleration of the urbanization process, urban subways are rapidly developed, and more underground base structures are built along with the urban subways. In recent years, tunnel safety problems, such as tunnel cracks, water leakage, settlement, lining spalling, chipping and the like, frequently occur, which cause serious casualties and huge economic losses, and thus tunnel safety is a key problem during tunnel operation.
The traditional manual inspection and closed circuit television system can cause serious consequences if the problems are not found timely, and the tunnel disease inspection is preferably carried out by combining manual and informatization means. Along with the continuous development of unmanned aerial vehicle technique, unmanned aerial vehicle application field is more and more extensive, has been taking photo by plane, agriculture, plant protection, express delivery transportation, disaster rescue, survey and drawing, electric power are patrolled and examined etc. and are using, and unmanned aerial vehicle is becoming the brand-new development direction of tunnel detection, along with the continuous innovation of unmanned aerial vehicle's location, perception, control and data transmission technique, future unmanned aerial vehicle will replace the mankind to go deep into secret danger area, explores the tunnel disease, guarantee infrastructure safety.
If the tunnel diseases are discovered in time, the normal use of the tunnel can be influenced, and even the structure safety is influenced. The traditional method for detecting by means of manpower and equipment consumes a large amount of manpower and material resources, has the problems of untimely detection, low detection efficiency and the like, and completely depends on subjective judgment of detection technicians.
Accordingly, the prior art is yet to be improved and developed.
Disclosure of Invention
The invention mainly aims to provide a tunnel detection method and a tunnel detection device based on an unmanned aerial vehicle, and aims to solve the problems that the traditional disease detection mainly depends on a visual method to preliminarily determine the disease type, the recognition rate depends on the experience of detection personnel, and the recognition rate is relatively high in uncertainty.
In order to achieve the above object, the present invention provides a tunnel detection method based on an unmanned aerial vehicle, which comprises the following steps:
receiving a video image of a tunnel acquired by an unmanned aerial vehicle;
preprocessing and standardizing the video image;
and inputting the video image after preprocessing and standardization processing into a trained three-dimensional convolution neural network model, and outputting a tunnel disease detection result by the three-dimensional convolution neural network model according to the video image.
Optionally, the method for detecting a tunnel based on a drone, where the receiving drone acquires video information of the tunnel, includes:
the unmanned aerial vehicle acquires image information of the surrounding environment in real time by carrying a laser radar or a thermal imaging high-definition camera;
acquiring pose information of the unmanned aerial vehicle through an inertial navigation device;
transmitting the image information and the pose information to a supercomputing platform in real time;
the super computing platform establishes a three-dimensional high-precision map according to the image information and the pose information, obtains a planned route and sends the planned route to the unmanned aerial vehicle;
and the unmanned aerial vehicle performs autonomous navigation and obstacle avoidance in tunnel flight according to the planned route.
Optionally, in the method for detecting a tunnel based on an unmanned aerial vehicle, a training process of the trained three-dimensional convolutional neural network model includes:
training the three-dimensional convolutional neural network model by applying a tunnel detection professional database;
acquiring a first percentage of video data in the application tunnel detection professional database for training a three-dimensional convolutional neural network model;
acquiring the video data of the second percentage left in the application tunnel detection professional database for verifying the trained three-dimensional convolution neural network model;
and when the recognition rate of the three-dimensional convolutional neural network model reaches a preset requirement, saving the current three-dimensional convolutional neural network model as a trained three-dimensional convolutional neural network model.
Optionally, the drone-based tunnel detection method wherein the first percentage is greater than the second percentage.
Optionally, the unmanned aerial vehicle-based tunnel detection method includes: and carrying out brightness adjustment, contrast adjustment and smooth filtering on the video image, wherein the smooth filtering is used for filtering sharp and discontinuous noise.
Optionally, in the method for detecting a tunnel based on an unmanned aerial vehicle, the video information input into the trained three-dimensional convolutional neural network model is a continuous multi-frame video image.
Optionally, in the method for detecting a tunnel based on a drone, the normalization process is to divide the continuous video into a number of short videos of 6 frames.
Optionally, in the unmanned aerial vehicle-based tunnel detection method, the three-dimensional convolutional neural network model is composed of an input layer, three convolutional layers, two sampling layers, a full connection layer, and an output layer.
Optionally, the unmanned aerial vehicle-based tunnel detection method includes: cracks, water leakage, settlement, steel bar exposure, slab staggering, lining stripping and chipping.
In addition, in order to achieve the above object, the present invention further provides a tunnel detection device based on an unmanned aerial vehicle, wherein the tunnel detection device based on an unmanned aerial vehicle includes:
the system comprises an unmanned aerial vehicle, an ultra-computation platform, an application tunnel detection professional database and a three-dimensional convolution neural network model;
the unmanned aerial vehicle is used for acquiring a video image of the tunnel;
the super-computation platform is used for establishing a three-dimensional high-precision map according to the image information acquired by the unmanned aerial vehicle and the pose information acquired by the inertial navigation device, and sending the map to the unmanned aerial vehicle after a planned route is obtained;
the application tunnel detection professional database is used for training the three-dimensional convolution neural network model;
and the three-dimensional convolutional neural network model is used for outputting a tunnel disease detection result according to the video image after training is finished.
The method comprises the steps of acquiring a video image of a tunnel by a receiving unmanned aerial vehicle; preprocessing and standardizing the video image; and inputting the video image after preprocessing and standardization processing into a trained three-dimensional convolution neural network model, and outputting a tunnel disease detection result by the three-dimensional convolution neural network model according to the video image. According to the invention, the video image of the tunnel is acquired by the unmanned aerial vehicle, and the video image is input into the trained three-dimensional convolutional neural network model, so that the characteristics of space and time in the video can be better captured, the tunnel disease detection result is rapidly output, the tunnel disease identification rate, the detection frequency and the processing speed are improved, and the artificial dependence degree is reduced.
Drawings
Fig. 1 is a flow chart of a preferred embodiment of the unmanned aerial vehicle-based tunnel detection method of the present invention;
fig. 2 is a schematic diagram of an implementation process of autonomous navigation and obstacle avoidance of the unmanned aerial vehicle in the preferred embodiment of the unmanned aerial vehicle-based tunnel detection method of the present invention;
FIG. 3 is a schematic diagram illustrating a 3D convolved convolution kernel performing a sliding window operation in a three-dimensional space of an input image according to a preferred embodiment of the unmanned aerial vehicle-based tunnel detection method of the present invention;
FIG. 4 is a schematic diagram of a three-dimensional convolutional neural network model according to a preferred embodiment of the unmanned aerial vehicle-based tunnel detection method of the present invention;
fig. 5 is a schematic diagram of a tunnel detection device based on an unmanned aerial vehicle according to a preferred embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer and clearer, the present invention is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
The convolutional neural network is a neural network method based on convolutional operation, belongs to one of representative networks of deep neural networks, has the characteristics of parameter sharing, local perception, multi-core performance and the like, and can effectively extract different characteristics by using a plurality of convolutional kernels for operation. A 3D-CNN is typically composed of a 3-dimensional convolution kernel, a two-dimensional Convolution (CNN) being a convolution in the spatial dimension, and a three-dimensional convolution being a convolution both in space and time. The CNN is widely applied to image identification, data information stored between continuous frames is not considered, the 3D-CNN can perform different convolution operations in the same area, different features can be extracted to perform different convolution operations, video data are analyzed, multi-channel information is generated from continuous multi-frame videos, convolution kernel sampling work is performed on each channel, and final feature representation is obtained through all channel information; compared with other neural networks, the convolutional nerves have excellent performance in an image understanding task, and particularly in the field of image classification, effective high-dimensional features can be automatically extracted.
The traditional method for detecting by means of manpower and equipment consumes a large amount of manpower and material resources, has the problems of untimely detection, low detection efficiency and the like, and completely depends on subjective judgment of detection technicians. The tunnel disease detection direction in the future should be automatic detection, highlighting the characteristics of modern technology automation and real-time. If the tunnel diseases are discovered in time, the normal use of the tunnel can be influenced, and even the structure safety is influenced. With the development of the technology, it is important to provide an intelligent tunnel detection method and an efficient detection system. A large number of video images can be obtained through the unmanned aerial vehicle, the unmanned aerial vehicle is combined with the big data cloud platform and the 3D-CNN, equipment integrating data collection, data storage and data processing is established, and the intelligent system equipment can be used for disease inspection of daily tunnels and is suitable for the development trend of future tunnel disease detection.
As shown in fig. 1, the tunnel detection method based on the unmanned aerial vehicle according to the preferred embodiment of the present invention includes the following steps:
and step S10, receiving the video image of the tunnel acquired by the unmanned aerial vehicle.
Specifically, the invention acquires the video image of the tunnel (the tunnel can also be an underground large-scale infrastructure) in real time through the unmanned aerial vehicle, the unmanned aerial vehicle has the functions of autonomous navigation and obstacle avoidance, the problem of autonomous navigation and obstacle avoidance of the unmanned aerial vehicle under the conditions of no GPS signal and insufficient light in the underground large-scale infrastructure such as the tunnel is solved, and the implementation process is as shown in fig. 2:
step A1: the unmanned aerial vehicle acquires image information of a surrounding environment in real time by carrying an OS-1 laser radar (a radar system for detecting characteristic quantities such as a position, a speed and the like of a target by emitting a laser beam, wherein the OS-1 laser radar is released by a laser radar manufacturer Ouster, the OS-1 can output a depth image, a signal image and an environment image with fixed resolution in real time, and a depth learning algorithm can utilize the data) or a thermal imaging high-definition camera (a sensor capable of detecting extremely small temperature difference and converting the temperature difference into a real-time video image to be displayed);
step A2: acquiring pose information of the unmanned aerial vehicle through an Inertial Measurement Unit (IMU), an Inertial navigation Unit (Inertial navigation Unit), and a sensor mainly used for detecting and measuring acceleration and rotary motion;
step A3: transmitting the environmental image information and the pose information around the unmanned aerial vehicle to a super computing platform in real time;
step A4: the supercomputing platform establishes a three-dimensional high-precision map by using an SLAM (Simultaneous Localization and Mapping, instant positioning and map construction, or synchronous positioning and map construction) to the surrounding environment image and the pose information of the unmanned aerial vehicle, performs route planning, and sends a planned route to the unmanned aerial vehicle;
step A5: the unmanned aerial vehicle receives the path planning information, and achieves autonomous navigation and obstacle avoidance of the unmanned aerial vehicle in the tunnel (the flight path is calculated by using a three-dimensional air route planning flight control algorithm, and autonomous navigation and obstacle avoidance of the unmanned aerial vehicle in the tunnel are achieved).
Further, after acquiring a video image of the tunnel, the unmanned aerial vehicle uploads the video image to a data storage platform such as hbase (hadoop database) to store mass data, so as to form a data warehouse of the large data platform, WiFi transmission is adopted for real-time data transmission, and when the tunnel has no WiFi signal, COFDM (coded Orthogonal Frequency Division multiplexing) technology can be adopted, and COFDM wireless real-time video transmission is realized by installing a transmitter and standard high-definition COFDM receiver equipment; the HBase data storage platform is a cloud database, adopts full-hosted NoSQL service of a cloud native and computing storage separation architecture, is compatible with various open-source standard interfaces such as HBase, Phoenix and OpenTSDB, is suitable for GB-PB level data storage, query and analysis and the like, and supports functions such as SQL analysis, secondary index, time sequence query and the like; the HBase is a distributed column-oriented database established on a Hadoop file system, and provides quick random access to massive structured data for a transverse development type database; high-quality analysis results can be obtained, and by establishing the database, big data support is provided for tunnel disease detection, and the objectivity of the detection is increased.
And step S20, preprocessing and standardizing the video image.
Specifically, the preprocessing includes brightness adjustment, contrast adjustment and smooth filtering of the image, so that the image features are obvious and easy to identify. Where the smoothing filter aims to remove sharp discontinuities in the noise. The normalization process is to divide the continuous video into several short 6-frame videos, since as the input window (time dimension of the convolution) increases, the number of trainable parameters also increases and the input to the 3D-CNN network is limited to a very small number of continuous video frames.
And step S30, inputting the video image after preprocessing and standardization processing into a trained three-dimensional convolution neural network model, wherein the three-dimensional convolution neural network model outputs a tunnel disease detection result according to the video image.
Specifically, the video information input into the trained three-dimensional convolutional neural network model is a continuous multi-frame video image, for example, the normalization process is to divide the continuous video into a plurality of 6-frame short videos, and then input into the trained three-dimensional convolutional neural network model.
The training process of the trained three-dimensional convolutional neural network model (namely, the 3D-CNN model) is as follows: training the three-dimensional convolutional neural network model by applying a tunnel detection professional database; acquiring a first percentage of video data in the application tunnel detection professional database for training a three-dimensional convolutional neural network model; acquiring the video data of the second percentage left in the application tunnel detection professional database for verifying the trained three-dimensional convolution neural network model; when the recognition rate of the three-dimensional convolutional neural network model reaches a preset requirement, the current three-dimensional convolutional neural network model is saved as a trained three-dimensional convolutional neural network model (the training process is a model autonomous learning process). The first percentage is greater than the second percentage, e.g., the first percentage is 70%, then the second percentage is 30%.
For example, a 3D-CNN model is trained by applying a tunnel detection professional database, 70% of data is used for training, 30% of data is used for verifying the model, the model is stored as the trained 3D-CNN model when the model reaches a high recognition rate, and the model has the capability of recognizing various diseases after being stored.
The three-dimensional convolutional neural network model outputs a tunnel disease detection result according to the video image, so that higher accuracy and more stable judgment can be obtained; the tunnel disease detection result comprises: cracks, water leakage, settlement, steel bar exposure, slab staggering, lining peeling and chipping, and of course, other disease results can also be obtained.
Fig. 3 is a schematic diagram showing a sliding window operation performed by a convolution kernel of 3D convolution in a three-dimensional space of an input image, where fig. 3 intends to illustrate that where 3D-CNN is different from 2D-CNN, 3D-CNN is based on 2D-CNN and performs sliding window in time series, and the convolution kernel moves 3D-CNN in three directions with respect to 2D-CNN, which is mainly different in that it can capture not only spatial information but also temporal information. Among them, the Sliding window (Sliding Windows) is called "Sliding window" for short, and is a work performed on the input layer; the input video image is traversed by designing the sliding window, the local image corresponding to each window is detected, the problem of input heterogeneity caused by scale, position, deformation and the like can be effectively solved, and the detection effect is improved.
The time characteristics in the video cannot be extracted in the prior art, and the 3D-CNN model obviously improves the extraction of the sample characteristics, can better capture the space and time characteristics in the video and improve the identification rate of tunnel diseases.
As shown in fig. 4, the three-dimensional convolutional neural network model (3D-CNN model) in the present invention is composed of an input layer, three convolutional layers, two sampling layers (i.e., pooling layers), a full-link layer, and an output layer.
The input layer is used for inputting video frame images of 6 continuous frames, and sliding window operation is carried out, so that the detection effect is improved.
Wherein, the convolutional layer is a 3D convolutional layer, first a series of small 3D feature extractors (kernel) need to be defined, the stacked high-level features are extracted, in order to generate a new feature space, different 3D kernel is used to extract different features in an input space, then a bias term is added, a linear activation function is used, and the formula is as follows:
wherein the content of the first and second substances,the ith 3D feature representing layer 1;a kth 3D feature space representing a previous layer;、、respectively representative valueIn the space of coordinate axesThe value of (1) is (b),representing three-dimensional feature extractor spaceCoordinates of (5);represents the activation function ReLU (rectified Linear Unit).
The pooling layer is a 3D maximum pooling layer, and is different from the correlation between input and kernel calculation in the convolutional layer, the pooling layer directly calculates the maximum value of elements in a pooling window, and the operation is called maximum pooling.
In the fully-connected layer, each neuron is connected with all neurons in the adjacent layer, before the fully-connected layer, firstly, a feature space needs to be flattened (flattened) to a neuron vector, then, vector-matrix multiplication is carried out, and a bias term is added and a linear activation function is applied.
Wherein the content of the first and second substances,for inputting feature vectors, from3D eigenspace of the layer (flatten);is the firstOutput feature vectors of layers (fully connected layers);representing a weight matrix;representing a bias term;representing the activation function ReLU.
Wherein the output layer adopts a Softmax function, the output of the Softmax function is a positive value between (0, 1), and the sum is 1; the output values of the multi-classification can be converted into probability distributions in the range of [0, 1] and 1 through a Softmax function; the calculation formula of the Softmax function is as follows:
wherein the content of the first and second substances,is the output value of the ith node; k is the number of output nodes, i.e. the number of classes classified.
The Softmax function is used in the multi-classification process, outputs the possibility of cracks, water leakage, sedimentation, steel bar exposure, slab staggering, lining stripping and block dropping to the interval of [0, 1], can be regarded as probability to understand, and finally selects the node with the maximum probability as an output result when the output node is selected finally.
According to the invention, a large number of video images can be obtained through the unmanned aerial vehicle, the unmanned aerial vehicle is combined with the big data cloud platform and the 3D-CNN, an equipment integrating data collection, data storage and data processing is established, and the intelligent system equipment can be used for disease inspection of daily tunnels and is suitable for the development trend of future tunnel disease detection.
Further, as shown in fig. 5, based on the above method for detecting a tunnel based on an unmanned aerial vehicle, the present invention also provides a device for detecting a tunnel based on an unmanned aerial vehicle, wherein the device for detecting a tunnel based on an unmanned aerial vehicle comprises:
the system comprises an unmanned aerial vehicle 100, a supercomputing platform 200, an application tunnel detection professional database 300 and a three-dimensional convolutional neural network model 400; the unmanned aerial vehicle 100 is used for acquiring a video image of a tunnel; the supercomputing platform 200 is configured to establish a three-dimensional high-precision map according to image information acquired by the unmanned aerial vehicle 100 and pose information acquired by an inertial navigation device, and send the map to the unmanned aerial vehicle 100 after obtaining a planned route; the application tunnel detection professional database 300 is used for training the three-dimensional convolutional neural network model 400; and the three-dimensional convolutional neural network model 400 is used for outputting a tunnel disease detection result according to the video image after training is completed.
The intelligent system equipment has the characteristics of automation, real-time performance, integration and the like, is efficient tunnel disease comprehensive detection system equipment, and improves the detection frequency, the processing speed and the manual dependence degree of tunnel diseases; the device can detect various diseases at one time, and belongs to comprehensive disease detection equipment; the labor cost is reduced.
In summary, the present invention provides a tunnel detection method and apparatus based on an unmanned aerial vehicle, the method includes: receiving a video image of a tunnel acquired by an unmanned aerial vehicle; preprocessing and standardizing the video image; and inputting the video image after preprocessing and standardization processing into a trained three-dimensional convolution neural network model, and outputting a tunnel disease detection result by the three-dimensional convolution neural network model according to the video image. According to the invention, the video image of the tunnel is acquired by the unmanned aerial vehicle, and the video image is input into the trained three-dimensional convolutional neural network model, so that the characteristics of space and time in the video can be better captured, the tunnel disease detection result is rapidly output, the tunnel disease identification rate, the detection frequency and the processing speed are improved, and the artificial dependence degree is reduced.
It should be noted that, in this document, 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 an … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
It is to be understood that the invention is not limited to the examples described above, but that modifications and variations may be effected thereto by those of ordinary skill in the art in light of the foregoing description, and that all such modifications and variations are intended to be within the scope of the invention as defined by the appended claims.
Claims (6)
1. A tunnel detection method based on an unmanned aerial vehicle is characterized by comprising the following steps:
receiving a video image of a tunnel acquired by an unmanned aerial vehicle;
receiving unmanned aerial vehicle acquires the video image in tunnel, still include before:
the unmanned aerial vehicle acquires image information of the surrounding environment in real time by carrying a laser radar or a thermal imaging high-definition camera;
acquiring pose information of the unmanned aerial vehicle through an inertial navigation device;
transmitting the image information and the pose information to a supercomputing platform in real time;
the super computing platform establishes a three-dimensional high-precision map according to the image information and the pose information, obtains a planned route and sends the planned route to the unmanned aerial vehicle;
the unmanned aerial vehicle performs autonomous navigation and obstacle avoidance in tunnel flight according to the planned route;
preprocessing and standardizing the video image; the pretreatment comprises the following steps: carrying out brightness adjustment, contrast adjustment and smooth filtering on the video image, wherein the smooth filtering is used for filtering sharp and discontinuous noise; the standardization processing is to divide the continuous video into a plurality of short videos with 6 frames;
inputting the video image after preprocessing and standardization processing into a trained three-dimensional convolutional neural network model, capturing the characteristics of space and time in the video, and outputting a tunnel disease detection result by the three-dimensional convolutional neural network model according to the video image;
the training process of the trained three-dimensional convolution neural network model comprises the following steps:
training the three-dimensional convolutional neural network model by applying a tunnel detection professional database;
acquiring a first percentage of video data in the application tunnel detection professional database for training a three-dimensional convolutional neural network model;
acquiring the video data of the second percentage left in the application tunnel detection professional database for verifying the trained three-dimensional convolution neural network model;
and when the recognition rate of the three-dimensional convolutional neural network model reaches a preset requirement, saving the current three-dimensional convolutional neural network model as a trained three-dimensional convolutional neural network model.
2. The drone-based tunnel detection method of claim 1, wherein the first percentage is greater than the second percentage.
3. The unmanned aerial vehicle-based tunnel detection method of claim 1, wherein the video images input into the trained three-dimensional convolutional neural network model are consecutive multi-frame video images.
4. The unmanned-aerial-vehicle-based tunnel detection method of claim 1, wherein the three-dimensional convolutional neural network model is composed of one input layer, three convolutional layers, two sampling layers, one fully-connected layer, and one output layer.
5. The unmanned aerial vehicle-based tunnel detection method of claim 1, wherein the tunnel defect detection result comprises: cracks, water leakage, settlement, steel bar exposure, slab staggering, lining stripping and chipping.
6. The utility model provides a tunnel detection device based on unmanned aerial vehicle, its characterized in that, tunnel detection device based on unmanned aerial vehicle includes:
the system comprises an unmanned aerial vehicle, an ultra-computation platform, an application tunnel detection professional database and a three-dimensional convolution neural network model;
the unmanned aerial vehicle is used for acquiring a video image of the tunnel; the unmanned aerial vehicle acquires image information of the surrounding environment in real time by carrying a laser radar or a thermal imaging high-definition camera; acquiring pose information of the unmanned aerial vehicle through an inertial navigation device; transmitting the image information and the pose information to a supercomputing platform in real time; the super computing platform establishes a three-dimensional high-precision map according to the image information and the pose information, obtains a planned route and sends the planned route to the unmanned aerial vehicle; the unmanned aerial vehicle performs autonomous navigation and obstacle avoidance in tunnel flight according to the planned route;
preprocessing and standardizing the video image; the pretreatment comprises the following steps: carrying out brightness adjustment, contrast adjustment and smooth filtering on the video image, wherein the smooth filtering is used for filtering sharp and discontinuous noise; the standardization processing is to divide the continuous video into a plurality of short videos with 6 frames;
the super-computation platform is used for establishing a three-dimensional high-precision map according to the image information acquired by the unmanned aerial vehicle and the pose information acquired by the inertial navigation device, and sending the map to the unmanned aerial vehicle after a planned route is obtained;
the application tunnel detection professional database is used for training the three-dimensional convolution neural network model;
the training process of the trained three-dimensional convolution neural network model comprises the following steps: training the three-dimensional convolutional neural network model by applying a tunnel detection professional database; acquiring a first percentage of video data in the application tunnel detection professional database for training a three-dimensional convolutional neural network model; acquiring the video data of the second percentage left in the application tunnel detection professional database for verifying the trained three-dimensional convolution neural network model; when the recognition rate of the three-dimensional convolutional neural network model reaches a preset requirement, saving the current three-dimensional convolutional neural network model as a trained three-dimensional convolutional neural network model;
and after the training of the three-dimensional convolutional neural network model is finished, capturing the characteristics of space and time in the video according to the video image, and outputting a tunnel disease detection result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110830687.1A CN113284144B (en) | 2021-07-22 | 2021-07-22 | Tunnel detection method and device based on unmanned aerial vehicle |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110830687.1A CN113284144B (en) | 2021-07-22 | 2021-07-22 | Tunnel detection method and device based on unmanned aerial vehicle |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113284144A CN113284144A (en) | 2021-08-20 |
CN113284144B true CN113284144B (en) | 2021-11-30 |
Family
ID=77286995
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110830687.1A Active CN113284144B (en) | 2021-07-22 | 2021-07-22 | Tunnel detection method and device based on unmanned aerial vehicle |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113284144B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114111725B (en) * | 2021-10-25 | 2024-02-20 | 中铁十六局集团第一工程有限公司 | Tunnel detection equipment |
CN114062516B (en) * | 2022-01-18 | 2022-09-27 | 中大检测(湖南)股份有限公司 | Tunnel-oriented ultrasonic detection system |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2017193331A (en) * | 2016-04-19 | 2017-10-26 | インダストリーネットワーク株式会社 | Drone flying object |
CN108764142A (en) * | 2018-05-25 | 2018-11-06 | 北京工业大学 | Unmanned plane image forest Smoke Detection based on 3DCNN and sorting technique |
CN109165562A (en) * | 2018-07-27 | 2019-01-08 | 深圳市商汤科技有限公司 | Training method, crosswise joint method, apparatus, equipment and the medium of neural network |
Family Cites Families (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104843176B (en) * | 2015-04-28 | 2017-04-19 | 北京恒华伟业科技股份有限公司 | Unmanned-gyroplane system used for automatic-inspection of bridges and tunnels and navigation method |
KR101794198B1 (en) * | 2016-10-26 | 2017-11-07 | 한국철도기술연구원 | Autonomous driving railway drone for mobilization to the scene of the fire |
US11159925B2 (en) * | 2018-05-08 | 2021-10-26 | Lg Electronics Inc. | Method and apparatus for transmitting V2X data in wireless communication system |
CN110456805B (en) * | 2019-06-24 | 2022-07-19 | 深圳慈航无人智能系统技术有限公司 | Intelligent tracking flight system and method for unmanned aerial vehicle |
CN111160301B (en) * | 2019-12-31 | 2023-04-18 | 同济大学 | Tunnel disease target intelligent identification and extraction method based on machine vision |
CN112598666A (en) * | 2021-01-06 | 2021-04-02 | 苏州光格科技股份有限公司 | Cable tunnel abnormity detection method based on convolutional neural network |
-
2021
- 2021-07-22 CN CN202110830687.1A patent/CN113284144B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2017193331A (en) * | 2016-04-19 | 2017-10-26 | インダストリーネットワーク株式会社 | Drone flying object |
CN108764142A (en) * | 2018-05-25 | 2018-11-06 | 北京工业大学 | Unmanned plane image forest Smoke Detection based on 3DCNN and sorting technique |
CN109165562A (en) * | 2018-07-27 | 2019-01-08 | 深圳市商汤科技有限公司 | Training method, crosswise joint method, apparatus, equipment and the medium of neural network |
Also Published As
Publication number | Publication date |
---|---|
CN113284144A (en) | 2021-08-20 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111862126B (en) | Non-cooperative target relative pose estimation method combining deep learning and geometric algorithm | |
CN106356757B (en) | A kind of power circuit unmanned plane method for inspecting based on human-eye visual characteristic | |
EP3633615A1 (en) | Deep learning network and average drift-based automatic vessel tracking method and system | |
KR101854554B1 (en) | Method, device and storage medium for calculating building height | |
CN112990310B (en) | Artificial intelligence system and method for serving electric robot | |
CN111126399B (en) | Image detection method, device and equipment and readable storage medium | |
CN111460968B (en) | Unmanned aerial vehicle identification and tracking method and device based on video | |
CN109255317B (en) | Aerial image difference detection method based on double networks | |
Yang et al. | Concrete defects inspection and 3D mapping using CityFlyer quadrotor robot | |
CN109598794B (en) | Construction method of three-dimensional GIS dynamic model | |
CN113284144B (en) | Tunnel detection method and device based on unmanned aerial vehicle | |
CN111985365A (en) | Straw burning monitoring method and system based on target detection technology | |
CN112507893A (en) | Distributed unsupervised pedestrian re-identification method based on edge calculation | |
CN113971666A (en) | Power transmission line machine inspection image self-adaptive identification method based on depth target detection | |
CN113936210A (en) | Anti-collision method for tower crane | |
CN113496163B (en) | Obstacle recognition method and device | |
CN117274627A (en) | Multi-temporal snow remote sensing image matching method and system based on image conversion | |
CN112465854A (en) | Unmanned aerial vehicle tracking method based on anchor-free detection algorithm | |
Li et al. | Driver drowsiness behavior detection and analysis using vision-based multimodal features for driving safety | |
CN116862832A (en) | Three-dimensional live-action model-based operator positioning method | |
CN110210436A (en) | A kind of vehicle-mounted camera line walking image-recognizing method | |
CN113920254B (en) | Monocular RGB (Red Green blue) -based indoor three-dimensional reconstruction method and system thereof | |
CN112069997B (en) | Unmanned aerial vehicle autonomous landing target extraction method and device based on DenseHR-Net | |
CN115731545A (en) | Cable tunnel inspection method and device based on fusion perception | |
CN117036721A (en) | Focusing method, apparatus, device, medium and program product for space station microscope |
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 |