CN113888462A - Crack identification method, system, readable medium and storage medium - Google Patents
Crack identification method, system, readable medium and storage medium Download PDFInfo
- Publication number
- CN113888462A CN113888462A CN202110994170.6A CN202110994170A CN113888462A CN 113888462 A CN113888462 A CN 113888462A CN 202110994170 A CN202110994170 A CN 202110994170A CN 113888462 A CN113888462 A CN 113888462A
- Authority
- CN
- China
- Prior art keywords
- picture
- crack
- detected
- subjected
- pixel blocks
- 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
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
- G06T7/0008—Industrial image inspection checking presence/absence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06T2207/10004—Still image; Photographic image
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 Analysis (AREA)
Abstract
The invention discloses a crack picture identification method, which is characterized in that a trained target detection algorithm is utilized to identify a picture to be detected to obtain a primary identification result, when the picture to be detected does not contain cracks according to the primary identification result, the picture to be detected is subjected to significance processing and then secondary identification, and a defective area is extracted, so that whether the image without the detected cracks is misjudged is further determined, and the reliability and the calculation speed of crack identification are increased.
Description
Technical Field
The invention relates to the technical field of picture identification, in particular to a crack identification method.
Background
With the great development of ultra/extra-high voltage power networks in recent years, the dry-type air-core reactor is used as important equipment for reactive compensation, and the purposes of providing reactive power for a system, improving power factor, reducing loss, improving electric energy quality, effectively controlling line distribution voltage and the like are achieved.
The dry-type air-core reactor has the following conditions when put into operation: under the working condition of full-load operation and the condition of frequent switching of the reactor, the heating expansion and cooling contraction of the conducting wire can be caused by the electrification and the power failure of the reactor. Secondly, the main component of the epoxy resin glass fiber reinforced plastic material layer encapsulated outside the reactor main body is a group with a benzene ring, and the material is easy to generate hydrogenation reaction and open the ring under the action of illumination. And thirdly, the outdoor working condition is affected by moisture, environment cold and heat, salt mist erosion and electromagnetic field all the year round. Under the comprehensive influence of the three conditions, the epoxy resin coated outside the dry-type reactor is easy to age, so that the mechanical strength and the insulation strength are rapidly reduced.
Since dry reactors are deployed in a relatively wide area, the manual validation approach is costly and inefficient, making it increasingly difficult to meet the demands of today's technological development.
Disclosure of Invention
The present invention is directed to solving at least one of the problems of the prior art. Therefore, the invention provides a crack identification method which can judge whether the picture to be identified contains cracks or not under the condition of various complex environments.
According to the crack picture identification method provided by the embodiment of the first aspect of the invention, the method comprises the following steps:
identifying the picture to be detected by utilizing a trained target detection algorithm to obtain a primary identification result;
when the picture to be detected does not contain cracks according to the primary identification result, performing visual saliency processing on the picture to be detected to obtain the picture subjected to the visual saliency processing;
identifying the picture subjected to the visual saliency processing to obtain a secondary identification result;
and determining whether the picture to be detected has cracks or not according to the secondary identification result.
According to some embodiments of the invention, the method further comprises:
collecting a plurality of crack pictures;
marking crack information of each crack picture in the collected crack pictures respectively, and forming a data set for training by the marked crack pictures;
and inputting the data set for training into a target detection algorithm for training to obtain the trained target detection algorithm.
According to some embodiments of the invention, the target detection algorithm is the YOLOv3 target detection algorithm.
According to some embodiments of the present invention, before labeling the crack information of each of the collected crack pictures, the method further includes:
changing the environment when the picture is shot to obtain the picture containing cracks under different environments; wherein the environment comprises at least one of a background, a brightness, an angle, a distance;
and/or the presence of a gas in the gas,
preprocessing the crack picture; wherein the preprocessing comprises at least one of cutting, amplifying, rotating and turning.
According to some embodiments of the invention, when the marking of the crack information is performed, the mark frame is made to accommodate the crack, and a minimum distance between the mark frame and an edge of the crack is smaller than a preset number of pixels.
According to some embodiments of the present invention, before the step of performing the visual saliency processing on the picture to be detected, the method further includes:
and removing redundant information in the picture to be detected by utilizing Gaussian difference calculation.
According to some embodiments of the present invention, before the step of performing the visual saliency processing on the picture to be detected, the method further includes:
and enhancing the edge detail information of the defect part in the picture to be detected by utilizing a multi-scale detail enhancement algorithm.
According to some embodiments of the present invention, the image to be detected is subjected to a visual saliency processing to obtain a visually saliency-processed image, including
Performing superpixel segmentation on a picture to be detected by a simple linear iterative clustering superpixel segmentation algorithm to obtain a superpixel set;
and performing visual saliency processing on the super-pixel set to obtain a picture subjected to the visual saliency processing.
According to some embodiments of the present invention, the super-pixel set includes a plurality of pixel blocks, and the step of identifying the picture subjected to the visual saliency processing to obtain the secondary identification result specifically includes:
determining the difference between all adjacent pixel blocks in the picture subjected to the saliency processing by comparing the change degrees of the colors between all adjacent pixel blocks in the picture subjected to the saliency processing;
extracting all pixel blocks with difference between the pixel blocks and the adjacent pixel blocks larger than a preset value to serve as target areas;
performing binary thresholding on the image subjected to the saliency processing according to the average gray value of the target region and the average gray values of other regions;
and identifying the image subjected to binary thresholding by using a contour detection algorithm to obtain a secondary identification result.
According to some embodiments of the present invention, the step of determining the difference between all adjacent pixel blocks in the saliency-processed picture by comparing the degree of change in color between all adjacent pixel blocks in the saliency-processed picture is specifically:
calculating the change degree of colors between all adjacent pixel blocks in the image subjected to the saliency processing under the multi-scale condition, and determining the difference under each scale according to the change degree of the colors;
and taking the average value of the differences of the adjacent pixel blocks under each scale as the difference between the adjacent pixel blocks.
A system for crack identification according to an embodiment of the second aspect of the invention is characterized by comprising:
the identification module can identify the picture to be detected by utilizing a trained target detection algorithm to obtain a primary identification result;
the saliency module is used for carrying out visual saliency processing on the picture to be detected when the picture to be detected does not contain cracks according to the primary recognition result to obtain the picture subjected to the visual saliency processing;
the defect extraction module is used for identifying the picture subjected to the visual saliency processing to obtain a secondary identification result;
and the crack judging module can determine whether the picture to be detected has cracks according to the secondary recognition result.
According to the computer readable medium of the third aspect of the invention, any one of the crack picture identification methods is realized.
A computer storage medium according to an embodiment of the fourth aspect of the invention, the storage medium having stored therein a computer program which, when executed, implements any of the above-described crack image identification methods.
According to the invention, the detection result is subjected to saliency processing on the basis of the existing target detection algorithm, and further identification is carried out according to the image subjected to saliency processing, so that the identification accuracy and efficiency can be further increased on the basis of the target detection algorithm.
Additional aspects and advantages of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.
Drawings
The above and/or additional aspects and advantages of the present invention will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:
FIG. 1 is a schematic diagram illustrating steps of a crack image recognition method according to an embodiment of the present invention;
FIG. 2 is a flowchart illustrating the training steps of the image recognition algorithm according to an embodiment of the present invention.
Detailed Description
Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the accompanying drawings are illustrative only for the purpose of explaining the present invention, and are not to be construed as limiting the present invention.
In the description of the present invention, it should be understood that the orientation or positional relationship referred to in the description of the orientation, such as the upper, lower, front, rear, left, right, etc., is based on the orientation or positional relationship shown in the drawings, and is only for convenience of description and simplification of description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus, should not be construed as limiting the present invention.
In the description of the present invention, the meaning of a plurality of means is one or more, the meaning of a plurality of means is two or more, and larger, smaller, larger, etc. are understood as excluding the number, and larger, smaller, inner, etc. are understood as including the number. If the first and second are described for the purpose of distinguishing technical features, they are not to be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated or implicitly indicating the precedence of the technical features indicated.
In the description of the present invention, unless otherwise explicitly limited, terms such as arrangement, installation, connection and the like should be understood in a broad sense, and those skilled in the art can reasonably determine the specific meanings of the above terms in the present invention in combination with the specific contents of the technical solutions.
With the great development of the ultra/extra-high voltage power grid, the dry-type air-core reactor is used as important equipment for reactive compensation, and plays an important role in a power system. In the actual application scene, due to the increase of the service time and the external abrasion, the material wrapped outside the dry-type air-core reactor main body is easy to crack, and the overall insulation and mechanical strength are further influenced.
Because the dry-type air reactor is quite wide in arrangement and the efficiency is too low by regularly sending people to confirm, a crack picture identification method is provided, and cracks can be identified by an algorithm in a camera monitoring mode.
The first embodiment,
On the premise of obtaining a target recognition algorithm which is trained, the crack recognition method described in the application comprises the following steps, referring to fig. 1:
b100, acquiring a picture to be detected, and carrying out standardized processing on the picture to be detected;
and arranging a camera beside the equipment needing to monitor the crack condition, and then transmitting the equipment state shot by the camera through a network to be used as a picture to be detected. For the YOLOv3 algorithm, the picture to be detected needs to be cut into 416 × 416 pixels, and the cut picture can obtain the best recognition effect in the YOLOv3 algorithm.
It is understood that the step of the present application B100 is a step based on YOLOv3, and if the target recognition algorithm is replaced, it is possible that the picture does not need to be standardized.
In the step B100, it is considered that one step is required in order to achieve a better recognition effect in the YOLOv3 algorithm, and if another target recognition algorithm is used, the purpose of the present application is not exceeded as long as the effect of recognizing cracks can be achieved.
B200, identifying the picture to be detected by using a trained target detection algorithm to obtain a primary identification result;
and inputting the standardized pictures into a target detection algorithm, and extracting all pictures containing cracks by the algorithm. And separating the picture with the crack from the picture without the crack as a preliminary identification result.
Although most of the pictures with cracks can be detected from the preliminary recognition result obtained in step B200, in order to further improve the performance of the crack recognition method, it is necessary to further process the pictures in which cracks are not detected, so as to increase the applicable range of the crack recognition method.
That is, after the step B200, if the image has a crack, the recognition process is ended, and if the image does not have a crack recognized, the following steps are performed.
And B300, judging whether the picture to be detected contains cracks or not.
And when the preliminary identification result determines that the picture to be detected does not contain cracks, entering the step B301, and performing visual saliency processing on the picture to be detected to obtain the picture subjected to the visual saliency processing.
It is conceivable that if a crack is found in the picture to be detected in the preliminary recognition result, the next step is not required.
Before the step of performing the visual saliency processing on the picture to be detected, the method further comprises the following steps of:
b301, removing redundant information in the picture to be detected by utilizing Gaussian difference calculation;
the frequency of the pictures is an index reflecting the degree of change in the gray-scale value. The low frequencies of the picture are contours and the high frequencies are noise and detail. The Difference of Gaussian (DOG) is the Difference of Gaussian functions, which can effectively suppress high frequency information by subtracting one picture from another to remove all redundant information except the frequencies that remain in the original picture.
Since the gaussian difference belongs to the prior art, the principle and formula thereof will not be explained in detail here.
In this embodiment, the method of removing noise in a picture by using a gaussian difference operator is merely an example, and it is not beyond the scope of knowledge of those skilled in the art to replace the method with another noise removal algorithm.
And step B302, enhancing the edge detail information of the defect part in the picture to be detected by utilizing a multi-scale detail enhancement algorithm.
While high-frequency noise is removed by Gaussian difference calculation, the information content of defect areas such as cracks is inevitably weakened. The edge detail information of the defective portion is thus enhanced by a multi-scale detail enhancement algorithm (multi-scale detail boosting).
Since the multi-scale detail enhancement algorithm belongs to the prior art, the principle thereof is not described in depth here, and only a simple explanation is made.
The core of the multi-scale detail enhancement algorithm is to filter the original image by using Gaussian kernels of three scales and then subtract the original image. The images of three different scales are respectively G1、G2、G3Image G1、G2、G3Respectively is σ1＝1、σ2＝2、σ3The 3 Blurred images (Blurred images) B were obtained by processing the images 4 using the following expression (1)1、B2、B3：
And integrating the 3 pictures with different detail degrees to obtain an overall detail picture. Wherein, the weights corresponding to the detail maps of the three scales are w respectively1＝0.5，w2＝0.5，w3＝0.25。
D*＝(1-w1×sgn(D1))D1+w2D2+w3D3 (2)
The use of multi-scale detail enhancement algorithms is well known in the art and will not be described in detail herein.
Through multi-scale enhancement calculation, details such as burr defect high-frequency noise and the like in the picture can be effectively reduced.
It can be understood that the multi-scale detail enhancement algorithm used in the present embodiment is a means for compensating the detail loss caused by denoising, and the objective is to improve the effect of the overall crack image identification method. It would also be within the spirit of the present application if the multi-scale enhancement algorithm used in the present embodiment were replaced with other types of detail-supplementing algorithms.
B400, performing superpixel segmentation on the picture to be detected to form a superpixel set by a simple linear iterative clustering superpixel segmentation algorithm; and taking the super-pixilated picture as a picture subjected to visual saliency processing.
The pictures acquired by the camera are high in resolution, so that the waiting time in the calculation process of the significance detection algorithm is long, and high hardware resources are consumed. Therefore, a processing mode for reducing the calculation amount and improving the significance detection efficiency is provided: the original picture is subjected to superpixel segmentation through a Simple Linear Iterative Clustering (SLIC) superpixel segmentation algorithm to form a series of superpixel sets.
The SLIC superpixel segmentation algorithm belongs to the prior art, and the specific process is not described herein again.
It is understood that the steps B301, B302, and B400 are steps for increasing the picture recognition speed and the picture recognition success rate, and the results can be normally obtained even if the steps are not performed, but the operation speed is slow. This example is presented to show a preferred embodiment.
B500, extracting the defect area of the image subjected to the saliency processing to obtain a secondary identification detection result;
the judgment of the salient region mainly follows the following principles: through comparison of color changes, the gray value area with large change corresponds to a higher significance value, and the uniform or fuzzy area has a low significance value; the crack and the background texture are different in gray scale characteristics, and the larger the inter-class variance between the crack and the background texture is, the larger the difference between two parts forming the picture is. The difference in grey value in the picture is used herein to mark crack significant areas. According to the analysis, the algorithm core should be to define the difference between the comparison pixel blocks, and the formula is shown in the following formula 3:
wherein d isg(pi，pj) Representing the difference in gray values between pixel block i and pixel block j; giAnd gjRepresenting the gray values of the pixel block i and the pixel block j respectively; dpos(pi，pj) Represents the distance between the pixel block i and the pixel block j, and mi ni and mj nj represent the horizontal and vertical coordinates of the central coordinate point of the pixel block i and the pixel block j, respectively.
When the variation tendency between pixel blocks is significant, the larger the value is found.
Preferably, as further improving the contrast between the salient region and other background non-salient regions, when calculating the difference between pixel blocks, multiple scale (scale) calculations need to be introduced, so that the salient region can be more obvious.
Equation 4 is a multiple scale calculation significance value equation, qkIs a pixel block piThe similar kth pixel block, r is a scale value. The comparison of the salient region with other regions in the picture is better improved by taking the average of the saliency values at multiple scales through equation 4.
And B600, performing binary thresholding on the image subjected to the saliency processing according to the average gray value of the target region and the average gray values of other regions.
If the value of the pixel disparity obtained in step B500 is larger than 65, the disparity between the two adjacent pixel blocks is considered to be large. And determining all pixel blocks with larger gray difference with the surrounding pixel blocks. These pixel blocks are taken as the key areas.
And determining the average gray value of the key region and the average gray value of other regions, taking the average of the average gray values as a threshold value for OpernCV to carry out binary thresholding on the image, and then carrying out binary thresholding on the image.
B700, identifying the image subjected to binary thresholding by utilizing a contour detection algorithm to obtain a secondary identification result
And extracting boundary information in the significant image subjected to binary thresholding by using a contour detection algorithm, marking a connected region, and framing and determining the position. And obtaining a secondary recognition result.
Preferably, findContours and drawContours functions in OpenCV2 can be used.
And B800, determining whether the picture to be detected has cracks or not according to the secondary identification result.
If the contour detection algorithm is positioned in a certain area in the image in the secondary identification result, the crack exists in the image; if the contour detection algorithm cannot be positioned to the middle of the graph but is positioned at the edge of the graph, the fact that no crack exists in the graph is indicated, and therefore whether the secondarily identified image has the crack or not is determined.
And integrating the picture with the crack in the primary recognition result and the picture with the crack area framed in the secondary recognition result to obtain all pictures with the crack.
The image recognition algorithm, the image processing algorithm and the contour detection algorithm are all the prior art, and in the field of image segmentation, different algorithms are usually suitable for different scenes, and many algorithms are improved based on the prior algorithms, so that the algorithms with similar use principles belong to the modes which can be thought by the skilled in the art.
Based on the same invention concept, the embodiment of the invention also provides a system for identifying the cracks, which comprises an identification module, a saliency module, a defect extraction module and a crack judgment module.
The recognition module is used for recognizing the standardized pictures by utilizing a trained target detection algorithm to obtain a primary recognition result; and if the picture is determined to contain cracks in the primary identification result, directly skipping other modules and outputting the result that the picture contains cracks.
The saliency module is used for carrying out visual saliency processing on the pictures without the detected cracks to obtain the pictures subjected to the visual saliency processing when the preliminary identification result obtained by the identification module determines that the pictures to be detected do not contain the cracks;
the defect extraction module is used for extracting the defect area of the image subjected to the visual saliency processing by utilizing a contour recognition algorithm to obtain a secondary recognition result;
and the crack judging module is used for determining whether the picture to be detected has cracks or not according to the secondary recognition result.
The specific principle is similar to the principle of the crack identification method described in the foregoing embodiments, and is usually implemented in the form of a computer program, which is not described herein again.
Example II,
In order to understand the contents described in the present application, the crack identification method described in the present application is based on an image detection algorithm, and the image detection algorithm needs to be trained. An example is provided here of how to train the image detection algorithm.
As shown in FIG. 2, an embodiment of the present invention provides a method of training an image recognition algorithm. The method comprises the steps of S100, S200 and S300.
In this embodiment, the image recognition algorithm based on the YOLOv3 neural network includes:
s100, collecting a plurality of crack pictures;
to obtain a data set for training the picture recognition algorithm, a large number of crack pictures need to be collected first. And finding the object with cracks in a manual inspection mode, and taking pictures for the cracks to obtain a plurality of crack pictures for training.
Since the effect of the picture recognition algorithm has a large correlation with the size of the training set itself. Two improvements have been proposed in order to increase the number of crack pictures used for training.
According to some preferred embodiments of the present application, step S100 collects several crack images, which may be specifically divided into:
step a, collecting a plurality of crack pictures.
And b, changing the environment when the picture is shot to obtain the picture containing cracks under different environments.
The above-mentioned environment when changing the picture of shooing mainly reflects in several items in background, luminance, shooting angle, the distance when changing the picture of shooing to prevent that the training set from appearing the bias.
And c, preprocessing the crack picture.
If the change of the shooting environment is not enough to obtain enough training sets, the pictures can be cut, turned, rotated and the like, so that the number of the pictures in the training sets is further increased.
On the basis of the crack pictures collected in the step a, the number of the crack pictures can be increased to more than 4 times of the original pictures at least after the processing of the steps b and c.
It should be noted that only one of the above steps b and c may exist, and the scheme described herein is a preferred embodiment.
Step S200, marking crack information on the collected crack pictures respectively to obtain a data set for training;
in the YOLOv3 picture recognition algorithm, we labeled the pictures using LabelImg software.
In actual operation, the size of the marking frame should be just enough to accommodate the crack. Because when the marking frame is too large, the pattern in the frame can contain too much background; when the marking frame is too small, the crack characteristics are the same as some impurity texture defects in the background after network processing, so that the model is difficult to identify after training. According to practical operation experience, the frame is closest to the crack image, and the distance between the frame and the crack image is less than 10 pixels, so that a better training model can be obtained in practice.
In addition, when a crack picture is selected, a picture with a larger difference between the foreground and the background is selected, so that the situation that the foreground and the background are mixed by a picture recognition algorithm to influence the effect of the picture recognition algorithm can be avoided.
According to practical use experience, the pictures collected by sorting are uniformly cut into 416 x 416 pixels, so that the recognition effect is the best. If a picture with large resolution exists, the picture with large resolution can be cut and then operated by program processing. The characteristics are determined according to the characteristics of the YOLOv3 picture recognition algorithm, and the specific reasons belong to the prior art and are not described herein again.
It is understood that steps S100 and S200 described in the present embodiment are based on the premise that a trained object detection algorithm is not available. This embodiment provides only a better training method, and a more complete and detailed explanation of the crack identification method described in this application is not meant to represent that this step is necessarily present. When the images are identified in batches, the training process of the algorithm also needs to be carried out once, and the trained algorithm can be used all the time.
S300, inputting a data set for training into a target detection algorithm for training to obtain a trained target detection algorithm;
the target detection algorithm used in this embodiment is a YOLOv3 neural network target detection algorithm, which belongs to the prior art that has been disclosed, and specific principles and operation modes are not described herein again.
Based on the same inventive concept, the embodiment of the present invention further provides a computer-readable medium and a storage medium, where the computer-readable medium and the storage medium contain program codes, and when the program codes are run on a computing device, the program codes are used for causing the computing device to execute the steps of the above crack picture identification method. Because the principle of solving the problem of the computer-readable medium and the storage medium is similar to that of the crack picture identification method, the implementation of the computer-readable medium and the storage medium can refer to the implementation of the picture identification method, and repeated details are not repeated.
The program product may employ any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium.
Program code for carrying out operations of the present invention may be written in any combination of one or more programming languages. The program code may execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device and partly on a remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computing device.
The embodiments of the present invention have been described in detail with reference to the accompanying drawings, but the present invention is not limited to the above embodiments, and various changes can be made within the knowledge of those skilled in the art without departing from the gist of the present invention.
Claims (13)
1. A crack identification method, characterized by comprising the steps of:
identifying the picture to be detected by utilizing a trained target detection algorithm to obtain a primary identification result;
when the picture to be detected does not contain cracks according to the primary identification result, performing visual saliency processing on the picture to be detected to obtain the picture subjected to the visual saliency processing;
identifying the picture subjected to the visual saliency processing to obtain a secondary identification result;
and determining whether the picture to be detected has cracks or not according to the secondary identification result.
2. The crack identification method of claim 1, further comprising,
collecting a plurality of crack pictures;
marking crack information of each crack picture in the collected crack pictures respectively, and forming a data set for training by the marked crack pictures;
and inputting the data set for training into a target detection algorithm for training to obtain the trained target detection algorithm.
3. The crack identification method of claim 1, wherein the target detection algorithm is the YOLOv3 target detection algorithm.
4. The crack identification method according to claim 2, wherein before labeling the crack information of each of the collected crack pictures, the method further comprises:
changing the environment when the picture is shot to obtain the picture containing cracks under different environments; wherein the environment comprises at least one of a background, a brightness, an angle, a distance;
and/or the presence of a gas in the gas,
preprocessing the crack picture; wherein the preprocessing comprises at least one of cutting, amplifying, rotating and turning.
5. The crack recognition method according to claim 2, wherein when the crack information is marked, the mark frame is made to accommodate the crack, and a minimum distance between the mark frame and an edge of the crack is smaller than a predetermined number of pixels.
6. The crack identification method according to claim 1, wherein before the step of performing the visual saliency processing on the picture to be detected to obtain the visually saliency-processed picture, the method further comprises:
and removing redundant information in the picture to be detected by utilizing Gaussian difference calculation.
7. The crack identification method according to any one of claims 1 to 6, wherein before the step of performing the visual saliency processing on the picture to be detected to obtain the visually saliency-processed picture, the method further comprises:
and enhancing the edge detail information of the defect part in the picture to be detected by utilizing a multi-scale detail enhancement algorithm.
8. The crack identification method according to claim 1, wherein the picture to be detected is subjected to visual saliency processing to obtain a picture subjected to visual saliency processing, and the picture comprises
Performing superpixel segmentation on a picture to be detected by a simple linear iterative clustering superpixel segmentation algorithm to obtain a superpixel set, and taking the superpixel set picture as a picture subjected to visual saliency processing.
9. Crack identification method as claimed in any of the claims 1-6 or 8, characterized in that the super-pixel set comprises a number of pixel blocks; the step of identifying the picture subjected to the visual saliency processing to obtain a secondary identification result specifically comprises the following steps:
determining the difference between all adjacent pixel blocks in the picture subjected to the saliency processing by comparing the change degrees of the colors between all adjacent pixel blocks in the picture subjected to the saliency processing;
extracting all pixel blocks with difference between the pixel blocks and the adjacent pixel blocks larger than a preset value to serve as target areas;
performing binary thresholding on the image subjected to the saliency processing according to the average gray value of the target region and the average gray values of other regions;
and identifying the image subjected to binary thresholding by using a contour detection algorithm to obtain a secondary identification result.
10. The crack identification method according to claim 9, wherein the step of determining the difference between all adjacent pixel blocks in the saliency-processed picture by comparing the degree of color variation between all adjacent pixel blocks in the saliency-processed picture comprises:
calculating the change degree of colors between all adjacent pixel blocks in the image subjected to the saliency processing under the multi-scale condition, and determining the difference under each scale according to the change degree of the colors;
and taking the average value of the differences of the adjacent pixel blocks under each scale as the difference between the adjacent pixel blocks.
11. A system for crack identification, comprising:
the identification module can identify the picture to be detected by utilizing a trained target detection algorithm to obtain a primary identification result;
the saliency module is used for carrying out visual saliency processing on the picture to be detected when the picture to be detected does not contain cracks according to the primary recognition result to obtain the picture subjected to the visual saliency processing;
the defect extraction module is used for identifying the picture subjected to the visual saliency processing to obtain a secondary identification result;
and the crack judging module can determine whether the picture to be detected has cracks according to the secondary recognition result.
12. A computer-readable medium, in which a computer program is stored which, when being executed by a processor, carries out a crack identification method as claimed in any one of claims 1 to 10.
13. A computer storage medium, characterized in that the storage medium has stored therein a computer program which, when executed, can implement the crack identification method described in any one of claims 1 to 10.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110994170.6A CN113888462A (en) | 2021-08-27 | 2021-08-27 | Crack identification method, system, readable medium and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110994170.6A CN113888462A (en) | 2021-08-27 | 2021-08-27 | Crack identification method, system, readable medium and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113888462A true CN113888462A (en) | 2022-01-04 |
Family
ID=79011248
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110994170.6A Pending CN113888462A (en) | 2021-08-27 | 2021-08-27 | Crack identification method, system, readable medium and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113888462A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115082479A (en) * | 2022-08-23 | 2022-09-20 | 启东凯顺机械制造有限公司 | Machine part fatigue crack identification method based on saliency characteristics |
| CN115657599A (en) * | 2022-12-28 | 2023-01-31 | 歌尔股份有限公司 | Laser cutting compensation method, device and equipment and storage medium |
| CN117095316A (en) * | 2023-10-18 | 2023-11-21 | 深圳市思友科技有限公司 | Road surface inspection method, device, equipment and readable storage medium |
-
2021
- 2021-08-27 CN CN202110994170.6A patent/CN113888462A/en active Pending
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115082479A (en) * | 2022-08-23 | 2022-09-20 | 启东凯顺机械制造有限公司 | Machine part fatigue crack identification method based on saliency characteristics |
| CN115657599A (en) * | 2022-12-28 | 2023-01-31 | 歌尔股份有限公司 | Laser cutting compensation method, device and equipment and storage medium |
| CN115657599B (en) * | 2022-12-28 | 2023-02-28 | 歌尔股份有限公司 | Laser cutting compensation method, device, equipment and storage medium |
| CN117095316A (en) * | 2023-10-18 | 2023-11-21 | 深圳市思友科技有限公司 | Road surface inspection method, device, equipment and readable storage medium |
| CN117095316B (en) * | 2023-10-18 | 2024-02-09 | 深圳市思友科技有限公司 | Road surface inspection method, device, equipment and readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111428748B (en) | HOG feature and SVM-based infrared image insulator identification detection method | |
| CN109118479B (en) | Capsule network-based insulator defect identification and positioning device and method | |
| CN113888462A (en) | Crack identification method, system, readable medium and storage medium | |
| CN109596634B (en) | Cable defect detection method and device, storage medium and processor | |
| CN111260616A (en) | Insulator crack detection method based on Canny operator two-dimensional threshold segmentation optimization | |
| CN111008961B (en) | Transmission line equipment defect detection method and system, equipment and medium thereof | |
| CN107679495B (en) | Detection method for movable engineering vehicles around power transmission line | |
| CN106408025B (en) | Aerial Images insulator classifying identification method based on image procossing | |
| CN110619623B (en) | Automatic identification method for heating of joint of power transformation equipment | |
| CN113902641B (en) | Data center hot zone judging method and system based on infrared image | |
| CN110751619A (en) | Insulator defect detection method | |
| CN108335294A (en) | The power distribution room abnormality image-recognizing method of complex condition | |
| CN108921172B (en) | Image processing device and method based on support vector machine | |
| CN112508019A (en) | GIS isolation/grounding switch state detection method and system based on image recognition | |
| CN114926392A (en) | Highlight region removing method based on image fusion | |
| CN113033451A (en) | Overhead line fault identification method and system based on deep learning | |
| CN116311201A (en) | Substation equipment state identification method and system based on image identification technology | |
| CN115272350A (en) | Method for detecting production quality of computer PCB mainboard | |
| CN105374029A (en) | Segmenting method and system of transformer substation equipment infrared image interest areas | |
| Niu et al. | Electrical equipment identification method with synthetic data using edge-oriented generative adversarial network | |
| CN111310899B (en) | Power defect identification method based on symbiotic relation and small sample learning | |
| CN116934696A (en) | Industrial PCB defect detection method and device based on YOLOv7-Tiny model improvement | |
| CN115690470A (en) | Method for identifying state of switch indicator and related product | |
| CN115619796A (en) | Method and device for obtaining photovoltaic module template and nonvolatile storage medium | |
| CN114639022A (en) | Switch cabinet on-off state identification method and system based on SUFR template matching |
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 |