CN113822891B - Tunnel disease detection method fusing laser point cloud and panoramic image - Google Patents

Tunnel disease detection method fusing laser point cloud and panoramic image Download PDF

Info

Publication number
CN113822891B
CN113822891B CN202111400964.1A CN202111400964A CN113822891B CN 113822891 B CN113822891 B CN 113822891B CN 202111400964 A CN202111400964 A CN 202111400964A CN 113822891 B CN113822891 B CN 113822891B
Authority
CN
China
Prior art keywords
image
point cloud
tunnel
point
laser
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
Application number
CN202111400964.1A
Other languages
Chinese (zh)
Other versions
CN113822891A (en
Inventor
李清泉
熊智敏
朱家松
朱松
元鹏鹏
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shenzhen Zhiyuan Space Innovation Technology Co ltd
Original Assignee
Shenzhen Zhiyuan Space Innovation Technology Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Shenzhen Zhiyuan Space Innovation Technology Co ltd filed Critical Shenzhen Zhiyuan Space Innovation Technology Co ltd
Priority to CN202111400964.1A priority Critical patent/CN113822891B/en
Publication of CN113822891A publication Critical patent/CN113822891A/en
Application granted granted Critical
Publication of CN113822891B publication Critical patent/CN113822891B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

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
    • G06T7/0004—Industrial image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/22—Matching criteria, e.g. proximity measures
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/253—Fusion techniques of extracted features
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/90—Determination of colour characteristics
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10028—Range image; Depth image; 3D point clouds

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Quality & Reliability (AREA)
  • Image Processing (AREA)

Abstract

The invention provides a tunnel disease detection method fusing a laser point cloud and a panoramic image, and relates to the technical field of tunnel detection.S 1, scanning the profile of the cross section of a tunnel through a laser scanner to obtain the point cloud characteristics of apparent diseases of the tunnel, and simultaneously recording the panoramic image of the water delivery tunnel through a panoramic camera to obtain the image characteristics of the apparent diseases of the tunnel; s2, extracting point cloud characteristics and image characteristics of the apparent tunnel diseases; s3, fusing the point cloud characteristics and the image characteristics, establishing a corresponding relation between each point in the point cloud and a pixel in the image, and transmitting the candidate disease area in the point cloud to the image; and S4, taking the candidate disease area in the image as a crack seed point, filling the crack seed point in the disease area and the interference area in the image through a growing and communicating algorithm, inputting the point cloud characteristics and the image characteristics of the disease area and the interference area into a classifier for classification and identification, and identifying the crack.

Description

Tunnel disease detection method fusing laser point cloud and panoramic image
Technical Field
The invention relates to the technical field of tunnel detection, in particular to a tunnel disease detection method fusing laser point cloud and panoramic images.
Background
The long-distance water delivery tunnel is a main building of diversion and regulation engineering, diseases and aging defects are inevitably generated due to rock stratum properties, temperature stress, water flow scouring and the like in the process of construction and use, the water delivery tunnel diseases affect the internal safety of the tunnel, the water delivery tunnel diseases are further aggravated along with the lapse of time, small cracks in the initial stage can be gradually developed and enlarged, if small cracks cannot be detected and repaired in time, accidents such as breakage and collapse of the water delivery tunnel, pipeline settlement, pressure pipeline explosion and the like can be caused, the stability and development of a city are threatened, and social hazards are particularly serious, so that the water delivery tunnel diseases need to be detected regularly to master the service state of the tunnel.
The detection time is limited, the operation must be suspended before the detection of the water delivery tunnel, the accumulated water in the tunnel is emptied, and the water cut-off maintenance time is determined by the reservoir capacity of the water-receiving urban reservoir; (2) the detection environment is severe, no lighting facilities are arranged in the tunnel, and after the tunnel is operated for a period of time, silt and aquatic organisms are attached to the wall of the tunnel, the bottom plate is wet and slippery, and the tunnel is difficult to walk; (3) the length of the tunnel is over-long, the diameter is large, the length of the water delivery tunnel is over-long and is fluctuated, so that the detection is more time-consuming and labor-consuming; (4) the entrance is less, has restricted the volume of water delivery tunnel disease detection equipment, and large-scale equipment is inconvenient to get into. The water delivery tunnel disease detection still adopts an artificial general survey method at present, and the characteristics of the water delivery tunnel determine that the process of the artificial general survey method is time-consuming, labor-consuming and poor in detection precision, cannot realize digital monitoring of the disease development process, cannot perform scientific analysis and research, and cannot meet the requirements of the existing long-distance water delivery tunnel disease detection in China.
At present, most of domestic and foreign detection and relatively mature methods for cracks of water delivery tunnels are based on image processing technology, and the cracks are identified and evaluated from the gray texture characteristics of the cracks, but the water delivery tunnel images usually have the problems of low contrast, uneven illumination, serious noise pollution and the like, so that the cracks cannot be effectively detected.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides a tunnel disease detection method fusing a laser point cloud and a panoramic image, so as to solve the technical problems.
The technical method for solving the technical problem is as follows: a tunnel disease detection method fusing laser point cloud and panoramic images is improved in that: the method comprises the following steps: s1, scanning the profile of the cross section of the tunnel through a laser scanner to obtain point cloud characteristics of the apparent diseases of the tunnel, and simultaneously recording a panoramic image of the water delivery tunnel through a panoramic camera to obtain image characteristics of the apparent diseases of the tunnel; s2, extracting the point cloud characteristics and the image characteristics of the apparent tunnel diseases, wherein the extracted point cloud characteristics are represented as follows:
Figure 737998DEST_PATH_IMAGE001
，
Figure 56984DEST_PATH_IMAGE002
is the characteristic of the point cloud,
Figure 162343DEST_PATH_IMAGE003
respectively scanning position, width, depth, reflection intensity and spatial relation of candidate disease points in the section, wherein L and S are respectively length and shape of the candidate disease area; the extracted image features are represented as:
Figure 908582DEST_PATH_IMAGE004
wherein
Figure 856815DEST_PATH_IMAGE005
G represents the image gray scale for the image feature; s3, fusing the point cloud characteristics and the image characteristics, establishing a corresponding relation between each point in the point cloud and a pixel in the image, and transmitting the candidate disease area in the point cloud to the image; s4, taking the candidate disease area in the image as a crack seed point, filling the crack seed point in the disease area and the interference area in the image through a growing and communicating algorithm, and inputting the point cloud characteristics and the image characteristics of the disease area and the interference area into a classifierAnd classifying and identifying to identify the cracks.
In the method, in the step S2, the image feature of the apparent tunnel disease includes a color feature.
In the above method, the step S2 of extracting the point cloud feature of the apparent tunnel disease includes the following steps:
s201, registering a scanning section with a standard section through scanning section correction and section filtering, and positioning candidate disease points on the continuous scanning section;
s202, clustering candidate disease points on the continuous scanning section into discrete candidate disease areas through point cloud clustering;
s203, extracting the characteristics of candidate disease points in the continuous scanning section, including the reflection intensity i, the position p, the width W, the depth D, the spatial relation SR and the length L and the shape S of the candidate disease area, wherein the extracted point cloud characteristics are represented as:
Figure 979492DEST_PATH_IMAGE006
，
Figure 939358DEST_PATH_IMAGE007
is a point cloud feature.
In the above method, in the step S201, the scanning section is corrected to the cross section of the tunnel by using a piecewise ellipse fitting algorithm.
In the above method, in step S201, the scanning cross-section filtering uses a wavelet transform algorithm to transform the scanning cross-section into a continuous radius distribution waveform, where high-frequency components on the radius represent outliers on the scanning cross-section.
In the above method, in step S201, the scanned cross section is registered with the standard cross section, and the scanned cross section is matched with the standard cross section by using an iterative closest point algorithm.
In the above method, the step S3 includes the following steps:
s31, calculating the cylindrical equidistant projection coordinates of the laser measuring point in the panoramic camera coordinate system as
Figure 997444DEST_PATH_IMAGE008
：
Figure 42760DEST_PATH_IMAGE009
The panoramic camera adopts a fisheye lens and a projection center of the panoramic camera
Figure 703549DEST_PATH_IMAGE010
As an origin, each discrete laser measuring point
Figure 517921DEST_PATH_IMAGE011
From the origin of coordinates
Figure 996176DEST_PATH_IMAGE012
Will form a corresponding three-dimensional vector ray and establish the origin
Figure 528788DEST_PATH_IMAGE012
As the center of a sphere, the
Figure 993268DEST_PATH_IMAGE013
Is a spherical model of radius, edge
Figure 927725DEST_PATH_IMAGE014
Direction measuring point by laser
Figure 62035DEST_PATH_IMAGE015
Projecting the image on a spherical model to form projection points under a spherical polar coordinate system
Figure 81943DEST_PATH_IMAGE016
Is established with
Figure 84534DEST_PATH_IMAGE013
For projection radius, by passing through the origin
Figure 873499DEST_PATH_IMAGE012
In the vertical direction ofA projection cylinder model with the line as the axis, wherein
Figure 711134DEST_PATH_IMAGE017
For measuring points by laser
Figure 218339DEST_PATH_IMAGE015
The polar coordinates of the projection point under the spherical coordinate system,
Figure 759042DEST_PATH_IMAGE018
for measuring points by laser
Figure 668092DEST_PATH_IMAGE015
Coordinates under the cylindrical equidistant projection coordinate system,
Figure 144204DEST_PATH_IMAGE019
for measuring points by laser
Figure 138705DEST_PATH_IMAGE015
From the origin of coordinates
Figure 483098DEST_PATH_IMAGE012
The distance of (d);
s32, calculating the cylindrical equidistant projection coordinates of the panoramic image points in the panoramic camera coordinate system
Figure 246655DEST_PATH_IMAGE020
Comprises the following steps:
Figure 408515DEST_PATH_IMAGE021
the panoramic camera adopts a fish-eye lens, and a fish-eye imaging model is equivalent to a hemisphere by using the projection center of the panoramic camera
Figure 624733DEST_PATH_IMAGE022
As an origin, establish the origin
Figure 772817DEST_PATH_IMAGE023
As the center of a sphere, the
Figure 390880DEST_PATH_IMAGE024
Is a spherical model of radius, and is used for image points under a spherical polar coordinate system
Figure 474374DEST_PATH_IMAGE025
Wherein
Figure 177888DEST_PATH_IMAGE026
Which represents the angle of incidence,
Figure 864084DEST_PATH_IMAGE027
representing the projection angle of the image point in the phase plane,
Figure 602233DEST_PATH_IMAGE028
the image height can be known according to the equidistant projection formula of the fisheye lens for the focal length of the fisheye lens
Figure 371475DEST_PATH_IMAGE029
；
And S33, solving the maximum value of mutual information through a mutual information method, calculating the corresponding relation between the laser measuring points and the panoramic image pixels, and transmitting the disease candidate area to the panoramic image.
In the above method, the step S33 includes the following steps:
s331, define
Figure 562285DEST_PATH_IMAGE030
And
Figure 52172DEST_PATH_IMAGE031
the mutual information MI of (A) is:
Figure 254614DEST_PATH_IMAGE032
laser measuring point containing position
Figure 804544DEST_PATH_IMAGE033
And intensity of reflection
Figure 217071DEST_PATH_IMAGE034
Two attributes, point cloud original reflection intensity
Figure 510649DEST_PATH_IMAGE035
After conversion, the image is expressed as
Figure 957811DEST_PATH_IMAGE036
The panoramic image point is expressed as
Figure 537697DEST_PATH_IMAGE037
，
Figure 703099DEST_PATH_IMAGE038
Is a point cloud reflection intensity point set
Figure 534789DEST_PATH_IMAGE039
The point (b) in (c) is,
Figure 102036DEST_PATH_IMAGE040
is a panoramic camera gray point set
Figure 603556DEST_PATH_IMAGE041
The point (b) in (c) is,
Figure 256254DEST_PATH_IMAGE042
is that
Figure 891635DEST_PATH_IMAGE039
And
Figure 47810DEST_PATH_IMAGE041
the joint distribution of (a) and (b),
Figure 235077DEST_PATH_IMAGE043
and
Figure 109492DEST_PATH_IMAGE044
are respectively as
Figure 548564DEST_PATH_IMAGE039
And
Figure 824825DEST_PATH_IMAGE041
the marginal distribution of (a) of (b),
Figure 933726DEST_PATH_IMAGE045
the micro-elements representing x are represented by,
Figure 295437DEST_PATH_IMAGE046
a infinitesimal representing y;
s332, obtaining
Figure 272620DEST_PATH_IMAGE047
At this time
Figure 403387DEST_PATH_IMAGE039
And
Figure 932458DEST_PATH_IMAGE041
geometric alignment;
s333, calculating the corresponding relation between the laser point cloud and the panoramic image;
and S334, transmitting the disease candidate area to the panoramic image.
In the above method, in the step S331, the
Figure 781465DEST_PATH_IMAGE039
And
Figure 296760DEST_PATH_IMAGE041
the joint distribution of (a) is calculated using a gaussian mixture model and a kernel density estimate.
In the above method, in the step S331, the
Figure 547613DEST_PATH_IMAGE039
And
Figure 998317DEST_PATH_IMAGE041
is calculated using an edge histogram of the overlapping portions of the two data sets。
The invention has the beneficial effects that: the geometrical characteristics and the appearance characteristics which can be used for identifying apparent diseases of the tunnel are extracted, the problem of multi-characteristic description of the diseases on the surface of the water delivery tunnel is solved, the reflection intensity image and the gray level image are matched based on mutual information by utilizing the similarity of the laser reflection intensity image and the gray level image, and the point cloud and the gray level image are fused by combining the mapping relation of the laser reflection intensity image and the laser point cloud, so that the geometrical characteristics and the appearance characteristics of cracks of the tunnel are fused; the problem of high false crack detection rate caused by uneven illumination, foreign matter interference, weak damage information and low contrast in the crack detection of the water-conveying tunnel of the two-dimensional image is solved by extracting the three-dimensional characteristics of the crack, and the accuracy and the efficiency of searching for crack seed points in the two-dimensional image are improved by positioning the crack seed points in the two-dimensional image by utilizing the three-dimensional characteristics; by growing and communicating the cracks in the two-dimensional image through the crack seed points in the three-dimensional point cloud, the switching from the low-precision three-dimensional point cloud to the high-precision two-dimensional image is realized, and the problem of low crack detection precision in the three-dimensional point cloud is solved; the method comprehensively utilizes the advantages of obvious crack characteristics in the three-dimensional point cloud and the advantages of high precision of the two-dimensional image, realizes the detection of the cracks on the surface of the water delivery tunnel, and particularly has more obvious advantages on the detection of low-contrast and fine cracks.
Drawings
FIG. 1 is a flow chart of a tunnel disease detection method fusing laser point cloud and panoramic image.
Detailed Description
The invention is further illustrated with reference to the following figures and examples.
The conception, the specific structure, and the technical effects produced by the present invention will be clearly and completely described below in conjunction with the embodiments and the accompanying drawings to fully understand the objects, the features, and the effects of the present invention. It is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all embodiments, and those skilled in the art can obtain other embodiments without inventive effort based on the embodiments of the present invention, and all embodiments are within the protection scope of the present invention. In addition, all the connection/connection relations referred to in the patent do not mean that the components are directly connected, but mean that a better connection structure can be formed by adding or reducing connection auxiliary components according to specific implementation conditions. All technical characteristics in the invention can be interactively combined on the premise of not conflicting with each other.
Referring to fig. 1, the tunnel disease detection method fusing a laser point cloud and a panoramic image of the invention includes the following steps:
s1, scanning the profile of the tunnel section through a laser scanner to obtain point cloud characteristics of the apparent diseases of the tunnel, recording a panoramic image of the water delivery tunnel through a panoramic camera to obtain image characteristics of the apparent diseases of the tunnel, and completing acquisition of original data of the tunnel;
s2, extracting the point cloud characteristics and the image characteristics of the apparent tunnel diseases, wherein the characteristic extraction is to extract proper characteristics as the basis of crack identification, and the extracted point cloud characteristics are expressed as follows:
Figure 334620DEST_PATH_IMAGE048
，
Figure 653606DEST_PATH_IMAGE049
is the characteristic of the point cloud,
Figure 758965DEST_PATH_IMAGE050
respectively scanning position, width, depth, reflection intensity and spatial relation of candidate disease points in the section, wherein L and S are respectively length and shape of the candidate disease area;
specifically, the method for extracting the point cloud characteristics of the apparent tunnel diseases comprises the following steps:
s201, registering the scanning section with the standard section through scanning section correction and section filtering, and positioning candidate disease points on the continuous scanning section;
specifically, the scanning section is corrected to the cross section of the tunnel by adopting a piecewise ellipse fitting algorithm; the scanning section filtering adopts a wavelet transform algorithm to transform the scanning section into a continuous radius distribution waveform, high-frequency components on the radius represent abnormal points on the scanning section, and the wavelet transform algorithm can quickly locate the abnormal points and filter the abnormal points; registering the scanned section with the standard section, and matching the scanned section with the standard section by adopting an iterative closest point algorithm so as to extract candidate disease points;
s202, clustering candidate disease points on the continuous scanning section into discrete candidate disease areas through point cloud clustering;
s203, extracting the characteristics of candidate disease points in the continuous scanning section, including the reflection intensity i, the position p, the width W, the depth D, the spatial relation SR and the length L and the shape S of the candidate disease area, wherein the extracted point cloud characteristics are represented as:
Figure 629838DEST_PATH_IMAGE006
，
Figure 453438DEST_PATH_IMAGE007
is a point cloud feature;
the common image features comprise color features, texture features, shape features, spatial relationship features and the like, wherein the shape features and the spatial relationship features are affected by a shooting angle and can be distorted, the texture features are not suitable for local parts, tunnel diseases are often areas with similar local features, therefore, the method selects to extract the color features from the image features of the apparent tunnel diseases, and the extracted image features are expressed as follows:
Figure 310535DEST_PATH_IMAGE004
wherein
Figure 270401DEST_PATH_IMAGE005
G represents the image gray scale for the image feature;
s3, fusing the point cloud characteristics and the image characteristics, establishing a corresponding relation between each point in the point cloud and a pixel in the image, and transmitting the candidate disease area in the point cloud to the image;
specifically, the method comprises the following steps:
s31, calculating the cylinder equidistance of the laser measuring point under the panoramic camera coordinate systemProjection coordinates are
Figure 62908DEST_PATH_IMAGE008
：
Figure 373803DEST_PATH_IMAGE009
The point cloud data is composed of a large number of laser measuring points and already has geographic coordinates, and the panoramic camera adopts a fisheye lens and uses the projection center of the panoramic camera
Figure 34592DEST_PATH_IMAGE010
As an origin, each discrete laser measuring point
Figure 848964DEST_PATH_IMAGE011
From the origin of coordinates
Figure 327219DEST_PATH_IMAGE051
Will form a corresponding three-dimensional vector ray and establish the origin
Figure 859831DEST_PATH_IMAGE012
As the center of a sphere, the
Figure 324311DEST_PATH_IMAGE013
Is a spherical model of radius, edge
Figure 993190DEST_PATH_IMAGE052
Direction measuring point by laser
Figure 393078DEST_PATH_IMAGE015
Projecting the image on a spherical model to form projection points under a spherical polar coordinate system
Figure 412987DEST_PATH_IMAGE053
Is established with
Figure 415578DEST_PATH_IMAGE013
For projection radius, by passing through the origin
Figure 204542DEST_PATH_IMAGE012
A projection cylinder model with the vertical direction straight line as the axis, wherein
Figure 296038DEST_PATH_IMAGE017
For measuring points by laser
Figure 537664DEST_PATH_IMAGE015
The polar coordinates of the projection point under the spherical coordinate system,
Figure 343946DEST_PATH_IMAGE018
for measuring points by laser
Figure 128362DEST_PATH_IMAGE015
Coordinates under the cylindrical equidistant projection coordinate system,
Figure 729107DEST_PATH_IMAGE019
for measuring points by laser
Figure 723608DEST_PATH_IMAGE015
From the origin of coordinates
Figure 68002DEST_PATH_IMAGE012
So as to realize the conversion from the laser point cloud coordinate to the cylindrical coordinate;
s32, calculating the cylindrical equidistant projection coordinates of the panoramic image points in the panoramic camera coordinate system
Figure 956192DEST_PATH_IMAGE020
Comprises the following steps:
Figure 727839DEST_PATH_IMAGE021
the panoramic camera adopts a fish-eye lens, and a fish-eye imaging model is equivalent to a hemisphere by using the projection center of the panoramic camera
Figure 209636DEST_PATH_IMAGE022
As an origin, is established withOrigin point
Figure 357721DEST_PATH_IMAGE023
As the center of a sphere, the
Figure 851150DEST_PATH_IMAGE024
Is a spherical model of radius, and is used for image points under a spherical polar coordinate system
Figure 59278DEST_PATH_IMAGE025
Wherein
Figure 762791DEST_PATH_IMAGE026
Which represents the angle of incidence,
Figure 448988DEST_PATH_IMAGE027
representing the projection angle of the image point in the phase plane,
Figure 311770DEST_PATH_IMAGE028
the image height can be known according to the equidistant projection formula of the fisheye lens for the focal length of the fisheye lens
Figure 425220DEST_PATH_IMAGE029
；
S33, solving the maximum value of mutual information through a mutual information method, calculating the corresponding relation between the laser measuring points and the panoramic image pixels, and transmitting the disease candidate area to the panoramic image;
specifically, the method comprises the following steps:
s331, define
Figure 616030DEST_PATH_IMAGE030
And
Figure 105917DEST_PATH_IMAGE031
the mutual information MI of (A) is:
Figure 573939DEST_PATH_IMAGE032
laser measuring point containing position
Figure 123869DEST_PATH_IMAGE033
And intensity of reflection
Figure 801975DEST_PATH_IMAGE034
The two attributes can be accurately related by coordinates, and according to step S31, the original reflection intensity of the point cloud
Figure 829973DEST_PATH_IMAGE035
After conversion, the image is expressed as
Figure 401769DEST_PATH_IMAGE036
The point cloud reflection intensity image and the camera gray level image of the same scene have high similarity, the panoramic image points are mainly gray level attribute panorama, and according to the step S32, the image points are represented as a gray level image under the cylindrical equidistant projection
Figure 122600DEST_PATH_IMAGE037
，
Figure 553582DEST_PATH_IMAGE038
Is a point cloud reflection intensity point set
Figure 260638DEST_PATH_IMAGE039
The point (b) in (c) is,
Figure 562306DEST_PATH_IMAGE040
is a panoramic camera gray point set
Figure 454039DEST_PATH_IMAGE041
The point (b) in (c) is,
Figure 106737DEST_PATH_IMAGE042
is that
Figure 601172DEST_PATH_IMAGE039
And
Figure 22926DEST_PATH_IMAGE041
the joint distribution of (a) and (b),
Figure 85560DEST_PATH_IMAGE043
and
Figure 959975DEST_PATH_IMAGE044
are respectively as
Figure 274413DEST_PATH_IMAGE039
And
Figure 550674DEST_PATH_IMAGE041
the marginal distribution of (a) of (b),
Figure 518630DEST_PATH_IMAGE045
the micro-elements representing x are represented by,
Figure 145920DEST_PATH_IMAGE046
a infinitesimal representing y; further, the
Figure 982158DEST_PATH_IMAGE039
And
Figure 378504DEST_PATH_IMAGE041
the joint distribution of (A) is calculated by adopting a Gaussian Mixture Model (GMM) and Kernel Density Estimation (KDE); further, the
Figure 782941DEST_PATH_IMAGE039
And
Figure 366369DEST_PATH_IMAGE041
is calculated using an edge histogram of the overlapping portions of the two data sets
S332, obtaining
Figure 22609DEST_PATH_IMAGE047
At this time
Figure 273462DEST_PATH_IMAGE039
And
Figure 848800DEST_PATH_IMAGE041
geometric alignment;
s333, calculating the corresponding relation between the laser point cloud and the panoramic image;
s334, transmitting the disease candidate area to the panoramic image;
when in use
Figure 185103DEST_PATH_IMAGE039
And
Figure 628723DEST_PATH_IMAGE041
when the two-dimensional alignment is geometrically aligned with each other,
Figure 734082DEST_PATH_IMAGE047
the maximum value will be reached, therefore, the maximum value of the mutual information is solved by the mutual information method using all pixels in the image to evaluate the statistical measure, at which point the set of points
Figure 480321DEST_PATH_IMAGE039
Sum point set
Figure 38342DEST_PATH_IMAGE041
Geometrically aligned under a cylindrical projection coordinate system and expressed as coordinates
Figure 36385DEST_PATH_IMAGE054
And
Figure 996250DEST_PATH_IMAGE055
alignment of, i.e. intensity of point cloud reflection
Figure 913391DEST_PATH_IMAGE036
And the gray level of the panoramic image
Figure 224286DEST_PATH_IMAGE037
Are mapped one by one, and
Figure 9709DEST_PATH_IMAGE036
is formed by
Figure 824081DEST_PATH_IMAGE035
So as to be converted into the light-emitting diode,
Figure 177702DEST_PATH_IMAGE035
and the point cloud position
Figure 710314DEST_PATH_IMAGE056
One-to-one mapping is carried out, and the gray scale of the panoramic image can be finally known
Figure 784581DEST_PATH_IMAGE037
And the point cloud location
Figure 719039DEST_PATH_IMAGE057
Mapping one by one so as to calculate the corresponding relation between the laser point cloud and the panoramic image, so that the disease candidate area is easy to locate due to the obvious spatial characteristics of the position attributes in the point cloud, and the disease candidate area is transmitted to the panoramic image by utilizing the mapping relation between the point cloud position and the gray level of the panoramic image;
s4, taking the candidate disease area in the image as a crack seed point, filling the crack seed point in the disease area and an interference area in the image through a growing and communicating algorithm, wherein the interference area is an area with similar characteristics to the crack, such as shadow, oil stain or a wire groove, and the like, inputting the point cloud characteristics and the image characteristics of the disease area and the interference area into a classifier for classification and identification, and finally accurately identifying the disease crack, so that the method has more remarkable advantages in detection of low contrast and fine cracks; further, the growing and communicating algorithm adopts a watershed algorithm; further, the classifier is a decision tree classifier.
The method extracts the geometric characteristics and the appearance characteristics which can be used for identifying apparent diseases of the tunnel, solves the problem of multi-characteristic description of the diseases on the surface of the water delivery tunnel, matches the reflection intensity image and the gray level image based on mutual information by utilizing the similarity of the laser reflection intensity image and the gray level image, and realizes the fusion of the point cloud and the gray level image by combining the mapping relation of the laser reflection intensity image and the laser point cloud, thereby realizing the fusion of the geometric characteristics and the appearance characteristics of the cracks of the tunnel; the problem of high false crack detection rate caused by uneven illumination, foreign matter interference, weak damage information and low contrast in the crack detection of the water-conveying tunnel of the two-dimensional image is solved by extracting the three-dimensional characteristics of the crack, and the accuracy and the efficiency of searching for crack seed points in the two-dimensional image are improved by positioning the crack seed points in the two-dimensional image by utilizing the three-dimensional characteristics; by growing and communicating the cracks in the two-dimensional image through the crack seed points in the three-dimensional point cloud, the switching from the low-precision three-dimensional point cloud to the high-precision two-dimensional image is realized, and the problem of low crack detection precision in the three-dimensional point cloud is solved; the method comprehensively utilizes the advantages of obvious crack characteristics in the three-dimensional point cloud and the advantages of high precision of the two-dimensional image, realizes the detection of the cracks on the surface of the water delivery tunnel, and particularly has more obvious advantages on the detection of low-contrast and fine cracks.
While the preferred embodiments of the present invention have been illustrated and described, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.

Claims (8)

1. A tunnel disease detection method fusing laser point cloud and panoramic images is characterized by comprising the following steps: the method comprises the following steps:
s1, scanning the profile of the cross section of the tunnel through a laser scanner to obtain point cloud characteristics of the apparent diseases of the tunnel, and simultaneously recording a panoramic image of the water delivery tunnel through a panoramic camera to obtain image characteristics of the apparent diseases of the tunnel;
s2, extracting the point cloud characteristics and the image characteristics of the apparent tunnel diseases, wherein the extracted point cloud characteristics are represented as follows:
Figure 585145DEST_PATH_IMAGE001
，
Figure 472329DEST_PATH_IMAGE002
is the characteristic of the point cloud,
Figure 295929DEST_PATH_IMAGE003
respectively scanning position, width, depth, reflection intensity and spatial relation of candidate disease points in the section, wherein L and S are respectively length and shape of the candidate disease area; the extracted image features are represented as:
Figure 153027DEST_PATH_IMAGE004
wherein
Figure 378472DEST_PATH_IMAGE005
G represents the image gray scale for the image feature;
s3, fusing the point cloud characteristics and the image characteristics, establishing a corresponding relation between each point in the point cloud and a pixel in the image, and transmitting the candidate disease area in the point cloud to the image; the step S3 includes the following steps:
s31, calculating the cylindrical equidistant projection coordinates of the laser measuring point in the panoramic camera coordinate system as
Figure 420246DEST_PATH_IMAGE006
：
Figure 465562DEST_PATH_IMAGE007
The panoramic camera adopts a fisheye lens and a projection center of the panoramic camera
Figure 126351DEST_PATH_IMAGE008
As an origin, each discrete laser measuring point
Figure 206302DEST_PATH_IMAGE009
From the origin of coordinates
Figure 169710DEST_PATH_IMAGE010
Will form a corresponding three-dimensional vector ray and establish the origin
Figure 702323DEST_PATH_IMAGE010
As the center of a sphere, the
Figure 166802DEST_PATH_IMAGE011
Is a spherical model of radius, edge
Figure 101260DEST_PATH_IMAGE012
Direction measuring point by laser
Figure 484837DEST_PATH_IMAGE013
Projecting the image on a spherical model to form projection points under a spherical polar coordinate system
Figure 504745DEST_PATH_IMAGE014
Is established with
Figure 772916DEST_PATH_IMAGE015
For projection radius, by passing through the origin
Figure 561880DEST_PATH_IMAGE010
A projection cylinder model with the vertical direction straight line as the axis, wherein
Figure 867091DEST_PATH_IMAGE016
For measuring points by laser
Figure 374295DEST_PATH_IMAGE013
The polar coordinates of the projection point under the spherical coordinate system,
Figure 180577DEST_PATH_IMAGE017
for measuring points by laser
Figure 824048DEST_PATH_IMAGE013
Coordinates under the cylindrical equidistant projection coordinate system,
Figure 815007DEST_PATH_IMAGE018
for measuring points by laser
Figure 543928DEST_PATH_IMAGE013
From the origin of coordinates
Figure 153901DEST_PATH_IMAGE010
The distance of (d);
s32, calculating the cylindrical equidistant projection coordinates of the panoramic image points in the panoramic camera coordinate system
Figure 651879DEST_PATH_IMAGE019
Comprises the following steps:
Figure 564471DEST_PATH_IMAGE020
the panoramic camera adopts a fish-eye lens, and a fish-eye imaging model is equivalent to a hemisphere by using the projection center of the panoramic camera
Figure 780689DEST_PATH_IMAGE021
As an origin, establish the origin
Figure 928773DEST_PATH_IMAGE022
As the center of a sphere, the
Figure 546837DEST_PATH_IMAGE023
Is a spherical model of radius, and is used for image points under a spherical polar coordinate system
Figure 879598DEST_PATH_IMAGE024
Wherein
Figure 583112DEST_PATH_IMAGE025
Which represents the angle of incidence,
Figure 534887DEST_PATH_IMAGE026
representing the projection angle of the image point in the phase plane,
Figure 7457DEST_PATH_IMAGE027
the image height can be known according to the equidistant projection formula of the fisheye lens for the focal length of the fisheye lens
Figure 261852DEST_PATH_IMAGE028
；
S33, solving the maximum value of mutual information through a mutual information method, calculating the corresponding relation between the laser measuring points and the panoramic image pixels, and transmitting the disease candidate area to the panoramic image; the step S33 includes the following steps:
s331, define
Figure 452662DEST_PATH_IMAGE029
And
Figure 942549DEST_PATH_IMAGE030
the mutual information MI of (A) is:
Figure 535204DEST_PATH_IMAGE031
laser measuring point containing position
Figure 944189DEST_PATH_IMAGE032
And intensity of reflection
Figure 622295DEST_PATH_IMAGE033
Two attributes, point cloud original reflection intensity
Figure 915873DEST_PATH_IMAGE034
After conversion, the image is expressed as
Figure 503980DEST_PATH_IMAGE035
The panoramic image point is expressed as
Figure 959232DEST_PATH_IMAGE036
，
Figure 859055DEST_PATH_IMAGE037
Is a point cloud reflection intensity point set
Figure 690745DEST_PATH_IMAGE038
The point (b) in (c) is,
Figure 257992DEST_PATH_IMAGE039
is a panoramic camera gray point set
Figure 104437DEST_PATH_IMAGE040
The point (b) in (c) is,
Figure 491556DEST_PATH_IMAGE041
is that
Figure 392516DEST_PATH_IMAGE038
And
Figure 689636DEST_PATH_IMAGE040
the joint distribution of (a) and (b),
Figure 752270DEST_PATH_IMAGE042
and
Figure 361106DEST_PATH_IMAGE043
are respectively as
Figure 534598DEST_PATH_IMAGE038
And
Figure 810859DEST_PATH_IMAGE040
the marginal distribution of (a) of (b),
Figure 169028DEST_PATH_IMAGE044
a micro element representing the number of X,
Figure 530739DEST_PATH_IMAGE045
micro representing yElement;
s332, obtaining
Figure 507922DEST_PATH_IMAGE046
At which I and G are geometrically aligned;
s333, calculating the corresponding relation between the laser point cloud and the panoramic image;
s334, transmitting the disease candidate area to the panoramic image;
and S4, taking the candidate disease area in the image as a crack seed point, filling the crack seed point in the disease area and the interference area in the image through a growing and communicating algorithm, inputting the point cloud characteristics and the image characteristics of the disease area and the interference area into a classifier for classification and identification, and identifying the crack.
2. The tunnel disease detection method fusing the laser point cloud and the panoramic image as claimed in claim 1, characterized in that: in the step S2, the image feature of the apparent tunnel disease includes a color feature.
3. The tunnel disease detection method fusing the laser point cloud and the panoramic image as claimed in claim 1, characterized in that: in the step S2, extracting point cloud features of the apparent tunnel disease includes the following steps:
s201, registering a scanning section with a standard section through scanning section correction and section filtering, and positioning candidate disease points on the continuous scanning section;
s202, clustering candidate disease points on the continuous scanning section into discrete candidate disease areas through point cloud clustering;
s203, extracting the characteristics of candidate disease points in the continuous scanning section, including the reflection intensity i, the position p, the width W, the depth D, the spatial relation SR and the length L and the shape S of the candidate disease area, wherein the extracted point cloud characteristics are represented as:
Figure 779635DEST_PATH_IMAGE047
，
Figure 184071DEST_PATH_IMAGE048
is a point cloud feature.
4. The tunnel disease detection method fusing the laser point cloud and the panoramic image as claimed in claim 3, characterized in that: in the step S201, the scanning section is corrected to the cross section of the tunnel by using a piecewise ellipse fitting algorithm.
5. The tunnel disease detection method fusing the laser point cloud and the panoramic image as claimed in claim 3, characterized in that: in step S201, the scanning cross-section filtering uses a wavelet transform algorithm to transform the scanning cross-section into a continuous radius distribution waveform, where high-frequency components on the radius represent abnormal points on the scanning cross-section.
6. The tunnel disease detection method fusing the laser point cloud and the panoramic image as claimed in claim 3, characterized in that: in step S201, the scanned cross section is registered with the standard cross section, and the scanned cross section is matched with the standard cross section by using an iterative closest point algorithm.
7. The tunnel disease detection method fusing the laser point cloud and the panoramic image as claimed in claim 1, characterized in that: in the step S331, the
Figure 33079DEST_PATH_IMAGE038
And
Figure 813953DEST_PATH_IMAGE040
the joint distribution of (a) is calculated using a gaussian mixture model and a kernel density estimate.
8. The tunnel disease detection method fusing the laser point cloud and the panoramic image as claimed in claim 1, characterized in that: in the step S331, the
Figure 533647DEST_PATH_IMAGE038
And
Figure 233619DEST_PATH_IMAGE040
is calculated using the edge histograms of the overlapping portions of the two data sets.
CN202111400964.1A 2021-11-24 2021-11-24 Tunnel disease detection method fusing laser point cloud and panoramic image Active CN113822891B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111400964.1A CN113822891B (en) 2021-11-24 2021-11-24 Tunnel disease detection method fusing laser point cloud and panoramic image

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111400964.1A CN113822891B (en) 2021-11-24 2021-11-24 Tunnel disease detection method fusing laser point cloud and panoramic image

Publications (2)

Publication Number Publication Date
CN113822891A CN113822891A (en) 2021-12-21
CN113822891B true CN113822891B (en) 2022-03-11

Family

ID=78918214

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111400964.1A Active CN113822891B (en) 2021-11-24 2021-11-24 Tunnel disease detection method fusing laser point cloud and panoramic image

Country Status (1)

Country Link
CN (1) CN113822891B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117541537B (en) * 2023-10-16 2024-05-24 江苏星湖科技有限公司 Space-time difference detection method and system based on all-scenic-spot cloud fusion technology

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112254637A (en) * 2020-10-13 2021-01-22 成都天佑智隧科技有限公司 Tunnel excavation surface scanning device and detection method based on various fusion data

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106053475B (en) * 2016-05-24 2018-10-23 浙江工业大学 Tunnel defect tunneling boring dynamic device for fast detecting based on active panoramic vision
CN107392944A (en) * 2017-08-07 2017-11-24 广东电网有限责任公司机巡作业中心 Full-view image and the method for registering and device for putting cloud
CN109060821B (en) * 2018-07-10 2021-05-04 深圳大学 Tunnel disease detection method and tunnel disease detection device based on laser detection
CN109060820B (en) * 2018-07-10 2021-04-13 深圳大学 Tunnel disease detection method and tunnel disease detection device based on laser detection
CN111855664B (en) * 2020-06-12 2023-04-07 山西省交通科技研发有限公司 Adjustable three-dimensional tunnel defect detection system

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112254637A (en) * 2020-10-13 2021-01-22 成都天佑智隧科技有限公司 Tunnel excavation surface scanning device and detection method based on various fusion data

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
Studies on automatic extraction of precise 3D laser measurement data in specific section of subway tunnel;Chao Tang等;《PROCEEDINGS OF SPIE》;20201105;第1-10页 *

Also Published As

Publication number Publication date
CN113822891A (en) 2021-12-21

Similar Documents

Publication Publication Date Title
CN109544679B (en) Three-dimensional reconstruction method for inner wall of pipeline
CN106327532B (en) A kind of three-dimensional registration method of single image
CN109285145B (en) Multi-standing tree height measuring method based on smart phone
CN109583284B (en) Urban high-rise building height extraction method and device based on high-resolution SAR image
CN107610164B (en) High-resolution four-number image registration method based on multi-feature mixing
CN111047695B (en) Method for extracting height spatial information and contour line of urban group
TW200929067A (en) 3D image detecting, editing and rebuilding system
CN107767456A (en) A kind of object dimensional method for reconstructing based on RGB D cameras
TW201022708A (en) Method of change detection for building models
US20220148304A1 (en) Systems and methods for automated detection of changes in extent of structures using imagery
JPWO2013002280A1 (en) Three-dimensional feature data generation device, three-dimensional feature data generation method, and three-dimensional feature data generation program
CN109472802A (en) A kind of surface grid model construction method constrained certainly based on edge feature
CN111462326A (en) Low-cost 360-degree panoramic video camera urban pipeline three-dimensional reconstruction method and system
CN113971768A (en) Unmanned aerial vehicle-based three-dimensional dynamic detection method for power transmission line illegal building
CN113822891B (en) Tunnel disease detection method fusing laser point cloud and panoramic image
CN111473776A (en) Landslide crack monitoring method based on single-image close-range photogrammetry
WO2020093631A1 (en) Antenna downtilt angle measurement method based on depth instance segmentation network
CN110988909A (en) TLS-based vegetation coverage determination method for sandy land vegetation in alpine and fragile areas
CN117036326A (en) Defect detection method based on multi-mode fusion
CN107977968B (en) Building layered detection method based on building shadow information mining
Li et al. A novel framework for urban change detection using VHR satellite images
Stent et al. An Image-Based System for Change Detection on Tunnel Linings.
CN113947724A (en) Automatic line icing thickness measuring method based on binocular vision
CN112037192A (en) Method for collecting burial depth information in town gas public pipeline installation process
CN115937673B (en) Geographic element rapid change discovery method based on mobile terminal photo

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