CN112508947A - Cable tunnel abnormity detection method - Google Patents
Cable tunnel abnormity detection method Download PDFInfo
- Publication number
- CN112508947A CN112508947A CN202011588195.8A CN202011588195A CN112508947A CN 112508947 A CN112508947 A CN 112508947A CN 202011588195 A CN202011588195 A CN 202011588195A CN 112508947 A CN112508947 A CN 112508947A
- Authority
- CN
- China
- Prior art keywords
- image
- images
- points
- feature points
- difference
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 12
- 239000011159 matrix material Substances 0.000 claims description 14
- 238000007689 inspection Methods 0.000 claims description 13
- 238000000034 method Methods 0.000 claims description 10
- 230000002159 abnormal effect Effects 0.000 claims description 7
- 238000006073 displacement reaction Methods 0.000 claims description 6
- 230000005856 abnormality Effects 0.000 claims description 3
- 230000004044 response Effects 0.000 claims description 3
- 230000001131 transforming effect Effects 0.000 claims description 3
- 239000013598 vector Substances 0.000 claims description 3
- 230000008859 change Effects 0.000 description 6
- 239000000284 extract Substances 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 210000001525 retina Anatomy 0.000 description 1
- 230000002441 reversible effect Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
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
- G06T7/001—Industrial image inspection using an image reference approach
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
Abstract
The invention provides a cable tunnel abnormity detection method, which comprises the steps of extracting SIFT feature points from a patrol image and a standard image respectively, calculating the relative translation and rotation angle of the two images through the extracted SIFT feature points, determining the overlapping area of the images, calculating a difference value aiming at the overlapping area, and judging that the difference value intensity and the area are larger than a specified threshold value as abnormity. And an SIFT algorithm is used for anomaly detection in the field of cable tunnel detection, so that the detection accuracy is improved.
Description
Technical Field
The invention relates to the field of cable tunnel detection, in particular to the field of cable tunnel abnormity detection.
Background
At present, when image abnormity analysis is carried out, an inspection robot needs to agree with an angle at a position of a picture to shoot, then normalization processing and historical inspection shot image registration are carried out, then two registered images are respectively continued to be subjected to region segmentation, a plurality of features of each region image are extracted, the plurality of features are fused, finally, the difference degree of the corresponding features of the two images is calculated, the difference degree is compared with a set threshold value, and whether the current inspection shot image is an abnormal image is judged.
Because there is a certain error (2-4 cm) in the positioning of the robot, the camera is not in a fixed position, and the camera angle has a certain rotation, and the camera is closer to the object to be photographed (0.5 m-2 m), and the parallax is larger, so that the object to be photographed twice will change in rotation, translation, scaling, etc., because the angle of the photographing position changes, the position of one point in the image at time 1 and the image at time 2 will change, that is, the position of the point in the image at time 1 in the image at time 2 will change, and therefore the local difference between the image at time 1 and the image at time 2 cannot be directly compared.
In addition, a phenomenon occurs in which an object or part of an object appearing at time 1 does not appear in the picture at time 2, and an object or part of an object not appearing at time 1 appears at time 2. The result is that foreign matter appears in most images, which is clearly contrary to reality.
Disclosure of Invention
In order to solve the above problems, the present invention provides a method for detecting an abnormality of a cable tunnel, which can accurately determine an abnormal condition of the cable tunnel.
In order to solve the above technical problem, an embodiment of the present invention provides a method for detecting an abnormality of a cable tunnel, which is characterized in that SIFT feature points are extracted from a patrol inspection image and a standard image, relative translation and rotation angles of the two images are calculated through the extracted SIFT feature points, an overlapping region of the images is determined, a difference value is calculated for the overlapping region, and the difference value is considered to be abnormal when the difference value is greater than a predetermined threshold value, and the method includes the following steps:
constructing a Gaussian scale space, namely:wherein σ is called a scale space factor, which is a standard deviation of gaussian normal distribution and reflects the degree of blurring of the image, the larger the value of σ is, the more blurred the image is, the larger the corresponding scale is, L (x, y, σ) is called a gaussian scale space of the image, G (x, y, σ) is a gaussian function of variable scale, and the value is:i (x, y) is an input image, and SIFT feature operators are used for respectively extracting feature points from the inspection image and the standard image;
(II) calculating matched points in the inspection image and the standard image, namely when the ratio of the minimum Euclidean distance between the 128-dimensional description vectors of the feature points of the two images to the next minimum Euclidean distance is smaller than a specified threshold value, the feature points are matched points;
and (III) determining the perspective relation between the inspection image and the standard image, wherein the relation between the two images is as follows: x1= H × x2, where x1 is the inspection image and x2 is the standard image, that is:wherein: (u 1, v 1) is a point in image x1, (u 2, v 2) is a point in image x2, H is a homography matrix, H is a 3 x 3 invertible matrix with 8 degrees of freedom, i.e.:h1, H2, H4 and H5 are rotation amounts and dimensions, H3 and H6 are horizontal displacement and vertical displacement respectively, H7 and H8 are deformation amounts in the horizontal direction and the vertical direction respectively, 4 pairs of matching points are found in an image x1 and an image x2, and then an H matrix can be solved;
and (IV) finding out the difference part of the two images, transforming the image x1 by using an H matrix, finally enabling the image x1 to be at the same position as the image x2, determining the overlapping part of the two images, and solving the difference value of the overlapping part of the two images to obtain the difference part of the two images, wherein the difference part exceeds a threshold value and can be judged to be abnormal.
Preferably, the detection of the feature points may use a differential gaussian scale space DoG.
Preferably, the extreme points of low contrast and unstable edge response points are deleted.
And finally, determining the main direction of the feature point, calculating the argument and the amplitude of the gradient of each pixel point in the field taking the feature point as the center and taking 3 multiplied by 1.5 sigma as the radius, and counting the argument of the gradient by using a histogram, wherein the direction corresponding to the highest peak in the histogram is the direction of the feature point.
The technical scheme of the invention has the following beneficial effects:
aiming at various conditions such as foreign matter leaving, foreign person entering and the like which may occur in a high-voltage cable tunnel, relevant management personnel can be accurately and timely identified and prompted to confirm, the confirmed data are divided into two types, one type is correct and has abnormity indeed, the data are classified into an image abnormity library, the other type is an image which has difference but does not belong to abnormity, for example, a new cable or equipment is added after construction, the data are classified into a standard image library, a system stored in the standard image library finds similar images again to be treated as normal images, and along with continuous accumulation of the two types of data, the identification accuracy can be continuously improved through continuous operation of the data.
Drawings
FIG. 1 is a diagram illustrating the steps of the present invention.
Detailed Description
To make the technical problems, technical solutions and advantages of the present invention more apparent, the following detailed description is given with reference to the accompanying drawings and specific embodiments, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, not all of them.
As shown in fig. 1, the invention is a cable tunnel anomaly detection method, which extracts SIFT feature points from a patrol image and a standard image, calculates relative translation and rotation angles of the two images through the extracted SIFT feature points, determines an overlapping region of the images, calculates a difference value for the overlapping region, and determines that the difference value is abnormal if the strength and the area of the difference value are greater than a specified threshold value, and comprises the following steps:
the SIFT feature operator respectively extracts feature points of an image x1 and an image x2, wherein x1 is a patrol inspection image, x2 is a standard image, the imaging process of a human on a retina when the human is far away from an object is simulated through the blurring degree of the image, the image is blurred when the human is near the object and the size of the image is larger, and the image is a Gaussian scale space. The convolution operation is carried out on the image and the Gaussian function to blur the image, images with different blur degrees can be obtained by using different Gaussian kernels, and the Gaussian scale space of one image can be obtained by convolution of the image and different Gauss:
sigma is called scale space factor, is the standard deviation of Gaussian normal distribution, reflects the degree of image blurring, and the larger the value is, the more blurred the image is, and the larger the corresponding scale is. L (x, y, σ) represents the gaussian scale space of the image. G (x, y, sigma) is a Gaussian function with variable scale and takes the value:i (x, y) is an input image, and the purpose of constructing a scale space is to detect feature points existing at different scales. In order to find the extreme point of the scale space, each pixel point is compared with all neighboring points of its image domain (same scale space) and scale domain (neighboring scale space), and when it is larger (or smaller) than all neighboring points, the point is the feature point.
And (II) calculating matched points in the image x1 and the image x2, wherein SIFT feature matching adopts a simple and effective method, namely when the ratio of the minimum Euclidean distance to the next minimum Euclidean distance between the 128-dimensional description vectors of the feature points of the two images is less than a certain threshold value, the feature pairs are considered to be matched pairs. The physical meaning of this ratio is the uniqueness of the matching pair, i.e. when the ratio is small, it indicates that a certain feature point in image 1 is very similar to a feature point in image 2.
And (iii) determining the perspective relationship between the two images, because the lens parameters of each shooting of the camera are consistent, and only the position and the angle are different between the two shots, so that the change between the image at the moment 1 and the image at the moment 2 conforms to the perspective change, namely x1= H x2, wherein x1 is the image at the moment 1, x2 is the image at the moment 2, H is a homography matrix, H is a 3 x 3 reversible matrix, and the degree of freedom is 8:
wherein: (u 1, v 1) is a point in the image x1, (u 2, v 2) is a point in the image x2, h1, h2, h4, h5 are rotation amounts and dimensions, h3 and h6 are horizontal displacement and vertical displacement, respectively, and h7 and h8 are deformation amounts in the horizontal and vertical directions, respectively. Solving the H matrix requires finding 4 pairs of matching points in image x1 and image x 2. Obtaining the H matrix can transform the image x1 to the view angle of the image x2, calculate the overlapping part of the image x1 and the image x2, and further compare the difference of the two images. The coordinates of the 8 optimal points are substituted into the formula x1= H x2, and the following equation set is obtained after expansion,
this is a nonlinear equation with 8 parameters that can be solved by direct linear transformation. Where m1, m2 … m8 are the required unknowns, i.e. the 8 parameters in the H matrix.
And (IV) finding out the difference part of the two images, transforming the image x1 by using an H matrix, finally enabling the image x1 to be at the same position as the image x2, determining the overlapping part of the two images, and then solving the difference value of the overlapping part of the two images to obtain the difference part of the two images.
In SIFT, different scale spaces are represented by using gaussian blur of different parameters, and the scale spaces are constructed to detect feature points existing in different scales, and a method for detecting the feature points is LoG (laplacian of gaussian), but the calculation amount of LoG is relatively large, and LoG (difference gaussian) can be used to approximate LoG, so extreme points are detected in the scale space of LoG. And deleting extreme points of low contrast and unstable edge response points.
And determining the main direction of the characteristic point, calculating the argument and the amplitude of the gradient of each pixel point in the field taking the characteristic point as the center and taking 3 multiplied by 1.5 sigma as the radius, and then counting the argument of the gradient by using a histogram. The horizontal axis of the histogram is the gradient direction, the vertical axis is the accumulated value of the gradient amplitude corresponding to the gradient direction, and the direction corresponding to the highest peak in the histogram is the direction of the feature point.
The above description is for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art should be able to cover the technical scope of the present invention, the technical solutions according to the present invention and the inventive concept thereof, with equivalent replacement or change, which are within the technical scope of the present invention.
Claims (4)
1. A cable tunnel abnormity detection method is characterized in that SIFT feature points are respectively extracted from a patrol image and a standard image, the relative translation and rotation angle of the two images are calculated through the extracted SIFT feature points, the overlapping area of the images is determined, a difference value is calculated aiming at the overlapping area, the difference value strength and the area which are larger than a specified threshold value can be regarded as abnormal, and the method comprises the following steps:
constructing a Gaussian scale space, namely:wherein σ is called a scale space factor, which is a standard deviation of gaussian normal distribution and reflects the degree of blurring of the image, the larger the value of σ is, the more blurred the image is, the larger the corresponding scale is, L (x, y, σ) is called a gaussian scale space of the image, G (x, y, σ) is a gaussian function of variable scale, and the value is:i (x, y) is an input image, and SIFT feature operators are used for respectively extracting feature points from the inspection image and the standard image;
(II) calculating matched points in the inspection image and the standard image, namely when the ratio of the minimum Euclidean distance between the 128-dimensional description vectors of the feature points of the two images to the next minimum Euclidean distance is smaller than a specified threshold value, the feature points are matched points;
and (III) determining the perspective relation between the inspection image and the standard image, wherein the relation between the two images is as follows: x1= H x2, whereinX1 is the inspection image, x2 is the standard image, namely:wherein: (u 1, v 1) is a point in image x1, (u 2, v 2) is a point in image x2, H is a homography matrix, H is a 3 x 3 invertible matrix with 8 degrees of freedom, i.e.:h1, H2, H4 and H5 are rotation amounts and dimensions, H3 and H6 are horizontal displacement and vertical displacement respectively, H7 and H8 are deformation amounts in the horizontal direction and the vertical direction respectively, 4 pairs of matching points are found in an image x1 and an image x2, and then an H matrix can be solved;
and (IV) finding out the difference part of the two images, transforming the image x1 by using an H matrix, finally enabling the image x1 to be at the same position as the image x2, determining the overlapping part of the two images, and solving the difference value of the overlapping part of the two images to obtain the difference part of the two images, wherein the difference part exceeds a threshold value and can be judged to be abnormal.
2. The method as claimed in claim 1, wherein the detection of the feature points uses a difference gaussian scale space DoG.
3. The method as claimed in claim 2, wherein the extreme points of low contrast and unstable edge response points are deleted.
4. The method for detecting the abnormality of the cable tunnel according to claim 1, wherein the main direction of the feature point is determined, the argument and the amplitude of the gradient of each pixel point are calculated in a domain with the feature point as a center and with a radius of 3 x 1.5 σ, the argument of the gradient is counted by using a histogram, and the direction corresponding to the highest peak in the histogram is the direction of the feature point.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011588195.8A CN112508947A (en) | 2020-12-29 | 2020-12-29 | Cable tunnel abnormity detection method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011588195.8A CN112508947A (en) | 2020-12-29 | 2020-12-29 | Cable tunnel abnormity detection method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112508947A true CN112508947A (en) | 2021-03-16 |
Family
ID=74951959
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011588195.8A Pending CN112508947A (en) | 2020-12-29 | 2020-12-29 | Cable tunnel abnormity detection method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112508947A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115376073A (en) * | 2022-10-24 | 2022-11-22 | 广东科凯达智能机器人有限公司 | Foreign matter detection method and system based on feature points |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101957325A (en) * | 2010-10-14 | 2011-01-26 | 山东鲁能智能技术有限公司 | Substation equipment appearance abnormality recognition method based on substation inspection robot |
CN103593832A (en) * | 2013-09-25 | 2014-02-19 | 重庆邮电大学 | Method for image mosaic based on feature detection operator of second order difference of Gaussian |
CN106951899A (en) * | 2017-02-24 | 2017-07-14 | 李刚毅 | Method for detecting abnormality based on image recognition |
CN108304883A (en) * | 2018-02-12 | 2018-07-20 | 西安电子科技大学 | Based on the SAR image matching process for improving SIFT |
CN109727239A (en) * | 2018-12-27 | 2019-05-07 | 北京航天福道高技术股份有限公司 | Based on SURF feature to the method for registering of inspection figure and reference map |
WO2019134327A1 (en) * | 2018-01-03 | 2019-07-11 | 东北大学 | Facial expression recognition feature extraction method employing edge detection and sift |
-
2020
- 2020-12-29 CN CN202011588195.8A patent/CN112508947A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101957325A (en) * | 2010-10-14 | 2011-01-26 | 山东鲁能智能技术有限公司 | Substation equipment appearance abnormality recognition method based on substation inspection robot |
CN103593832A (en) * | 2013-09-25 | 2014-02-19 | 重庆邮电大学 | Method for image mosaic based on feature detection operator of second order difference of Gaussian |
CN106951899A (en) * | 2017-02-24 | 2017-07-14 | 李刚毅 | Method for detecting abnormality based on image recognition |
WO2019134327A1 (en) * | 2018-01-03 | 2019-07-11 | 东北大学 | Facial expression recognition feature extraction method employing edge detection and sift |
CN108304883A (en) * | 2018-02-12 | 2018-07-20 | 西安电子科技大学 | Based on the SAR image matching process for improving SIFT |
CN109727239A (en) * | 2018-12-27 | 2019-05-07 | 北京航天福道高技术股份有限公司 | Based on SURF feature to the method for registering of inspection figure and reference map |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115376073A (en) * | 2022-10-24 | 2022-11-22 | 广东科凯达智能机器人有限公司 | Foreign matter detection method and system based on feature points |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108932496B (en) | Method and device for counting number of target objects in area | |
CN108230237B (en) | Multispectral image reconstruction method for electrical equipment online detection | |
Klippenstein et al. | Quantitative evaluation of feature extractors for visual slam | |
Placht et al. | Rochade: Robust checkerboard advanced detection for camera calibration | |
EP2079054B1 (en) | Detection of blobs in images | |
CN112927249A (en) | System and method for finding image central line by visual system | |
US9767383B2 (en) | Method and apparatus for detecting incorrect associations between keypoints of a first image and keypoints of a second image | |
TWI639136B (en) | Real-time video stitching method | |
CN112017223A (en) | Heterologous image registration method based on improved SIFT-Delaunay | |
CN111563896A (en) | Image processing method for catenary anomaly detection | |
CN114331879A (en) | Visible light and infrared image registration method for equalized second-order gradient histogram descriptor | |
CN110222661A (en) | It is a kind of for motion estimate and the feature extracting method of tracking | |
CN112508947A (en) | Cable tunnel abnormity detection method | |
CN117314986A (en) | Unmanned aerial vehicle cross-mode power distribution equipment inspection image registration method based on semantic segmentation | |
Chen et al. | On registration of regions of interest (ROI) in video sequences | |
Heimbach et al. | Improving object tracking accuracy in video sequences subject to noise and occlusion impediments by combining feature tracking with Kalman filtering | |
Liu et al. | Outdoor camera calibration method for a GPS & camera based surveillance system | |
CN113706635B (en) | Long-focus camera calibration method based on point feature and line feature fusion | |
CN116206139A (en) | Unmanned aerial vehicle image upscaling matching method based on local self-convolution | |
CN113407030B (en) | Visual positioning method, related device, equipment and storage medium | |
US11645827B2 (en) | Detection method and device for assembly body multi-view change based on feature matching | |
Jun-bo | Template matching algorithm based on gradient search | |
CN113450335A (en) | Road edge detection method, road edge detection device and road surface construction vehicle | |
Zhou et al. | Speeded-up robust features based moving object detection on shaky video | |
Wang et al. | Application of improved SURF algorithm in real scene matching and recognition |
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 |