Image-based crack detection
Contact-less remote-sensing crack detection and/quantification methodologies are described, which are based on three-dimensional (3D) scene reconstruction, image processing, and pattern recognition. The systems and methodologies can utilize depth perception for detecting and/or quantifying cracks. These methodologies can provide the ability to analyze images captured from any distance and using any focal length or resolution. This adaptive feature may be especially useful for incorporation into mobile systems, such as unmanned aerial vehicles (UAV) or mobile autonomous or semi-autonomous robotic systems such as wheel-based or track-based radio controlled robots, as utilizing such structural inspection methods onto those mobile platforms may allow inaccessible regions to be properly inspected for cracks.
1. A system for crack detection, the system comprising:
a storage device; and
a processing system connected to the storage device; and
a program stored in the storage device, wherein execution of the program by the processing system causes the system to perform functions, including functions that:
(i) establish an appropriate structuring element based on a working distance and focal length of a three-dimensional (3D) structure of a scene; from a plurality of images of the scene;
(ii) segment potential crack patterns by applying a morphological operation;
(iii) determine appropriate features for each segmented pattern; and
(iv) classify a crack from a non-crack pattern using a trained classifier, thereby forming a multiscale crack map,
wherein:
for the plurality of images of the scene there are a plurality of data acquisition parameters, wherein the relation between different image acquisition parameters is in accordance with the following:
SF
=
(
WD
FL
)
(
SS
SR
)
n
,
where feature size (SF) is the size of a crack thickness represented by n pixels in an image, working distance (WD) is the working distance between camera and object, focal length (FL) is the camera focal length, sensor size (SS) (mm) is the camera sensor size, and sensor resolution (SR) is the camera sensor resolution in pixels; or
the multiscale crack map is formulated in accordance with the following:
and 0 otherwise;
where J m is the crack map at scale m of a structuring element, S min , is the minimum structuring element size, C k is the binary crack image obtained by using k as the structuring element, and u and v are the pixel coordinates of the crack map image.
2. The system of claim 1 , wherein the trained classifier comprises a neural network.
3. The system of claim 1 , wherein the trained classifier comprises a support vector machine (SVM).
4. The system of claim 1 , wherein the trained classifier comprises a neural nearest-neighbor classifier.
5. The system of claim 1 , wherein the morphological operation is substantially in accordance with the following:
T =max└( I∘S {0°,45°,90°,135°} ●S {0°,45°,90°,135°} ,I┘−I,
wherein I is the grayscale image, S is the structuring element that defines which neighboring pixels are included in the operation, ‘∘’ is the morphological opening, and ‘●’ is the morphological closing.
6. The system of claim 5 , wherein execution of the program by the processor further causes the system to perform a function that: eliminates extracted patterns if the length of a segmented pattern is less than a minimum length specified by a user.
7. The system of claim 5 , wherein execution of the program by the processor further causes the system to perform a function that: converts minimum length of interest in unit length to minimum length in pixels in accordance with the following:
l
p
=
(
FL
WD
)
l
,
wherein l is the defined length by the user in unit length, focal length (FL) is in pixels and working distance (WD) is in unit length, and l p is the length in pixels.
8. The system of claim 1 , wherein for the plurality of images of the scene there are a plurality of data acquisition parameters, wherein the relation between different image acquisition parameters is in accordance with the following:
SF
=
(
WD
FL
)
(
SS
SR
)
n
,
where feature size (SF) is the size of a crack thickness represented by n pixels in an image, working distance (WD) is the working distance between camera and object, focal length (FL) is the camera focal length, sensor size (SS) (mm) is the camera sensor size, and sensor resolution (SR) is the camera sensor resolution in pixels.
9. The system of claim 8 , wherein
SF
=
(
WD
FL
)
n
,
wherein FL is in pixels.
10. The system of claim 1 , further wherein the multiscale crack map is formulated in accordance with the following:
J m ( u,v )={1 ,∃kε[S min ,m];C k ( u,v )=1, and 0 otherwise;
where J m is the crack map at scale m of a structuring element, S min , is the minimum structuring element size, C k is the binary crack image obtained by using k as the structuring element, and u and v are the pixel coordinates of the crack map image.
11. A method of crack detection, the method comprising:
establishing an appropriate structuring element based on the working distance and focal length of a view for a scene;
segmenting potential crack patterns by applying a morphological operation;
determining appropriate features for each segmented pattern; and
classifying a crack from a non-crack pattern using a trained classifier; and
forming a multiscale crack map,
wherein:
the relation between different image acquisition parameters is in accordance with the following:
SF
=
(
WD
FL
)
(
SS
SR
)
n
,
where feature size (SF) is the size of a crack thickness represented by n pixels in an image, working distance (WD) is the working distance between camera and object, focal length (FL) is the camera focal length, sensor size (SS) (mm) is the camera sensor size, and sensor resolution (SR) is the camera sensor resolution in pixels; or
the multiscale crack map is formulated according to the following:
J m ( u,v )={1 ,∃kε[S min ,m];C k ( u,v )=1, and 0 otherwise;
where J m is the crack map at scale m of a structuring element, S min , is the minimum structuring element size, C k is the binary crack image obtained by using k as the structuring element, and u and v are the pixel coordinates of the crack map image.
12. The method claim 11 , wherein the trained classifier comprises a neural network.
13. The method of claim 11 , wherein the trained classifier comprises a support vector machine (SVM).
14. The method of claim 11 , wherein the trained classifier comprises a neural nearest-neighbor classifier.
15. The method of claim 11 , wherein the morphological operation is in accordance with the following:
T =max└( I∘S {0°,45°,90°,135°} ●S {0°,45°,90°,135°} ,I┘−I,
wherein I is the grayscale image, S is the structuring element that defines which neighboring pixels are included in the operation, ‘∘’ is the morphological opening, and ‘●’ is the morphological closing.
16. The method of claim 15 , wherein execution of the program by the processor further configures the system to perform a function to: eliminate extracted patterns if the length of a segmented pattern is less than a minimum length specified by a user.
17. The method of claim 15 , wherein execution of the program by the processor further configures the system to perform a function to: convert minimum length of interest in unit length to minimum length in pixels in accordance with the following:
l
p
=
(
FL
WD
)
l
,
wherein l is the defined length by the user in unit length, focal length (FL) is in pixels and working distance (WD) is in unit length, and l p is the length in pixels.
18. The method of claim 11 , wherein the relation between different image acquisition parameters is in accordance with the following:
SF
=
(
WD
FL
)
(
SS
SR
)
n
,
where feature size (SF) is the size of a crack thickness represented by n pixels in an image, working distance (WD) is the working distance between camera and object, focal length (FL) is the camera focal length, sensor size (SS) (mm) is the camera sensor size, and sensor resolution (SR) is the camera sensor resolution in pixels.
19. The method of claim 18 , wherein
SF
=
(
WD
FL
)
n
,
wherein FL is in pixels.
20. The method of claim 11 , further wherein the multiscale crack map is formulated according to the following:
J m ( u,v )={1 ,∃kε[S min ,m];C k ( u,v )=1, and 0 otherwise;
where J m is the crack map at scale m of a structuring element, S min , is the minimum structuring element size, C k is the binary crack image obtained by using k as the structuring element, and u and v are the pixel coordinates of the crack map image.