CN116433659A - Three-section road defect image processing method - Google Patents
Three-section road defect image processing method Download PDFInfo
- Publication number
- CN116433659A CN116433659A CN202310677633.5A CN202310677633A CN116433659A CN 116433659 A CN116433659 A CN 116433659A CN 202310677633 A CN202310677633 A CN 202310677633A CN 116433659 A CN116433659 A CN 116433659A
- Authority
- CN
- China
- Prior art keywords
- defect
- large frame
- frame
- small
- road
- 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.)
- Granted
Links
- 230000007547 defect Effects 0.000 title claims abstract description 136
- 238000003672 processing method Methods 0.000 title claims abstract description 20
- 230000002950 deficient Effects 0.000 claims abstract description 19
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 13
- 238000001514 detection method Methods 0.000 claims description 23
- 238000013135 deep learning Methods 0.000 claims description 7
- 238000005457 optimization Methods 0.000 claims description 6
- 230000007797 corrosion Effects 0.000 claims description 4
- 238000005260 corrosion Methods 0.000 claims description 4
- 230000006870 function Effects 0.000 claims description 4
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 238000000034 method Methods 0.000 abstract description 6
- 238000012545 processing Methods 0.000 abstract description 3
- 238000012549 training Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 3
- 230000008439 repair process Effects 0.000 description 3
- 238000013145 classification model Methods 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000007635 classification algorithm Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000011160 research Methods 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration using local operators
- G06T5/30—Erosion or dilatation, e.g. thinning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30132—Masonry; Concrete
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/06—Recognition of objects for industrial automation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Multimedia (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a three-section road defect image processing method, which relates to the field of image processing, has a simple model, is easy to realize, reduces the omission ratio, greatly improves the model precision, and meets the requirement of fine road management. The method specifically comprises the following steps: step 1: collecting road images with defects; step 2: detecting the position and defect type of a road defect in a road image; step 3: for any defect, cutting through a rectangular large frame to generate a large frame image containing a type of defect; step 4: cutting any large frame image into a plurality of rectangular small frames, and judging whether the two classes of the small frames are defects or non-defects; step 5: judging whether the two classification categories of the small frame are defects or not by optimizing and adjusting based on a bridge connection algorithm; step 6: the collection of all small frames in the large frame image, which are judged to be defective by all the two classification categories, is recorded as a defect image.
Description
Technical Field
The invention relates to the field of image processing, in particular to a three-section road defect image processing method.
Background
The influence of factors such as severe weather, natural disasters, traffic and the like, particularly the increase of traffic volume of heavy vehicles, has a great influence on the service life of roads along with the change of time. Over time, cracks, pits and other damages to the road surface can occur, the integrity of the road surface can be damaged, and the service cycle is shortened. It is important to find out the road defect problem in time to repair.
Early damage detection is mainly accomplished manually. However, due to limitations of manpower, financial resources, natural conditions, etc., and the subjectivity of manual detection is too strong, it is impractical to rely on only manual detection of defects. The existing research shows that the problem of pavement defect detection is solved by utilizing a deep learning algorithm. However, the general detection model is used for detecting road defects at present, and the following problems exist: (1) Road defect problems are various, the types of defects comprise 11 defects of cracks, blocky cracks, longitudinal cracks, transverse cracks, subsidence, ruts, wave hugs, pits, looseness, oil flooding and repair, the traditional analysis model does not consider the difference of each defect, all pictures are subjected to deep learning together, whether the defects are detected by only judging the defects through a threshold value, the detection precision is low, and the requirements of road defect detection in actual work cannot be met. (2) Through traditional threshold judgment, the problem of missing detection of defects with small partial pixel difference exists. (3) The road defect detection results at the present stage are rough, and the actual engineering needs finer management, such as knowing the area size of the defect, and further evaluating the severity of the defect.
Disclosure of Invention
The invention aims at solving the problems in the prior art, and designing a three-section road defect image processing method, wherein the defects and positions of a model are detected through a traditional model in the first stage, a large frame is intercepted in the second stage, each large frame image only comprises one defect, the large frame is divided into a defect small frame and a non-defect small frame in the third stage, and whether the classification type of the small frame is the defect or not is judged based on the optimization adjustment of a bridge connection algorithm.
A three-section road defect image processing method comprises the following steps:
step 1: collecting road images with defects;
step 2: detecting the position and defect type of a road defect in a road image;
step 3: for any defect, cutting through a rectangular large frame to generate a large frame image containing a type of defect;
step 4: cutting any large frame image into a plurality of rectangular small frames, and judging whether the two classes of the small frames are defects or non-defects;
step 5: judging whether the two classification categories of the small frame are defects or not by optimizing and adjusting based on a bridge connection algorithm;
step 6: the collection of all small frames in the large frame image, which are judged to be defective by all the two classification categories, is recorded as a defect image.
Further, step 5 includes the steps of:
step 5.1: for any large frame image, finding out all the defect small frames which are classified into the two categories and are judged to be defects;
step 5.2: collecting a plurality of defect small frames of the same category adjacent to each other to construct an island;
step 5.3: respectively calculating the shortest distance between two islands;
step 5.4: and when the shortest distance is smaller than or equal to the distance threshold value, bridging the two islands, and updating the two classification categories of the connected small frames into defects.
Preferably, the two island bridge connection comprises the following steps:
step 5.4.1: firstly, for two islands with the shortest distance less than or equal to a distance threshold D, finding out end points L at two ends of the shortest distance 1 And L 2 The coordinates are respectively denoted as (x) 1 ,y 1 ) And (x) 2 ,y 2 ）；
Drawn through L 1 And L 2 Is expressed as:
wherein x and y are the abscissa and ordinate of any point on the line, respectively;
finding out two classification non-defective small frames, and calculating the distance d from the non-defective small frames m to the straight line m :
Wherein d m Representing the distance from the non-defective small frame m to a straight line, x m And y m Respectively representing the abscissa and the ordinate of the non-defective small frame m, wherein A, B and C are calculated using the following formulas, respectively:
step 5.4.2: distance d from non-defective small frame m to straight line m And when the class adjustment threshold value d is less than or equal to, the non-defect small frame is used as a bridge, and the two classes are adjusted to be defects.
Further, step 3 optimizes the length and width of the large frame based on the small frame size:
wherein,,
to optimize the length of the rear large frame +.>
In order to optimize the width of the large frame, a is the length of the small frame, b is the width of the small frame, l is the initial length of the large frame image, d is the initial width of the large frame image, int is a rounding function, when decimal points exist, the part after the decimal points is directly removed,
and then the original large frame is replaced by the large frame based on the length and the width of the large frame after optimization, and the defect part is still contained in the new large frame.
Preferably, the coordinates are kept unchanged at any angle of the initial large frame to optimize
And->
And drawing the optimized large frame as the length and the width.
Further, solving the area of the road defect based on the defect image, wherein the specific formula is as follows:
wherein S is the defect area, a is the length of the small frame, b is the width of the small frame, n is the total number of the small frames with defects, and alpha is the conversion coefficient of the image length and the actual length.
Preferably, step 3 performs corrosion and expansion treatment on the large frame image to remove noise in the image, so that road defects in the large frame are more easily distinguished from background edges.
Further, step 2 detects defect categories and locations based on the deep learning object detection model.
Preferably, step 3 outlines the large frame image by a rectangular frame.
Preferably, the step 4 rectangular box size is 97×97 pixels.
Compared with the prior art, the invention has the beneficial effects that:
the invention aims at solving the problems in the prior art, and designing a three-section road defect image processing method, wherein the defects and positions of a model are detected through a traditional model in the first stage, a large frame is intercepted in the second stage, each large frame image only comprises one defect, the large frame is divided into a defect small frame and a non-defect small frame in the third stage, and whether the classification type of the small frame is the defect or not is judged based on the optimization adjustment of a bridge connection algorithm.
1. The invention creatively provides a three-stage thought, disassembles complex tasks into a plurality of single tasks, and the model training is easier, simple and efficient, and is suitable for various road defects. Compared with the traditional road defect detection, the method has the advantages that the deep learning multi-classification detection model is used, the road defect type and position are directly obtained, and the detection precision is greatly improved.
2. The invention creatively proposes a bridge connection algorithm, and effectively solves the problem of missed detection in threshold judgment by processing two short-distance island bridges.
3. According to the invention, the large frame is expanded through the size of the small frame, so that the expanded large frame is just used for accommodating an integer number of small frames, and the operation is simple and practical.
4. The invention carries out corrosion and expansion treatment on the large frame image, removes noise in the image, and ensures that road defects in the large frame are more easily distinguished from background edges.
5. The invention can solve the actual defect area of the road based on the number of the defect small frames, meets the requirement of road management refinement, and is convenient for the grading of the post-road management maintenance.
6. The invention decouples the reasoning model (namely the bridge classification algorithm) from the classification model, the parameters of the models are not shared, one model is not influenced when being optimized and the parameters of the other model are adjusted, and the fault tolerance of the model is high.
Drawings
FIG. 1 is a schematic flow chart of the present invention;
FIG. 2 is a large block diagram image view of the present invention;
FIG. 3 is a small block diagram image view of the present invention;
FIG. 4 is a schematic diagram of a bridge connection according to the present invention;
fig. 5 is a prior art deep learning model flow.
Detailed Description
The invention relates to a three-section road defect image processing method, which is further described in detail below with reference to the accompanying drawings and the specific implementation method.
The invention provides a three-section road defect image processing method, as shown in fig. 1, comprising the following steps:
step 1: collecting road images with defects;
step 2: detecting the position and defect type of a road defect in a road image;
step 3: for any defect, cutting through a rectangular large frame to generate a large frame image containing a type of defect, as shown in fig. 2;
step 4: for any large frame image, cutting into a plurality of rectangular small frames, and judging whether the small frames are classified as defects or not based on the classification judgment of the softmax classifier as shown in figure 3;
step 5: judging whether the two classification categories of the small frame are defects or not by optimizing and adjusting based on a bridge connection algorithm;
step 6: the collection of all small frames in the large frame image, which are judged to be defective by all the two classification categories, is recorded as a defect image.
The bridge connection algorithm has wide application range, and is suitable for various model defects such as cracks, blocky cracks, longitudinal cracks, transverse cracks, subsidence, ruts, wave hugs, pits, looseness, oil flooding, repair and the like.
The quality of the road image dataset of step 1 has great influence on the performance and accuracy of the model. Therefore, the data set tested by the invention contains enough samples, is representative, and is independent of each other. In particular, the image dataset is divided into two folders, one containing images with road defects and the other containing tag files. Meanwhile, the names of the images and the corresponding labels are consistent except for suffixes, so that the associated inquiry is facilitated.
It is particularly noted here that only one defect is contained in one large frame image, which is an important element that can streamline the model.
Further, step 2 detects defect categories and locations based on the deep learning object detection model. A classical deep learning object detection model is selected, which can be YOLO series or RCNN series. And then, according to specific tasks, parameters such as the number of categories in the model, a loss function and the like and a network structure are adjusted, and after the adjustment is finished, an initial learning rate and an optimizer are set, so that training can be started. In the training process, the model adjusts model parameters through gradient descent to obtain a model with minimum loss. And testing by using a testing set, wherein the evaluation parameters meet the requirements, and the trained weights are obtained. Otherwise, modifying the model parameters and the structure, and continuing training. The training flow of the model is shown in fig. 5.
Preferably, step 3 performs corrosion and expansion treatment on the large frame image to remove noise in the image, so that road defects in the large frame are more easily distinguished from background edges. And 3, drawing a large frame image through a rectangular frame. The manual direct sketching can be adopted, and the direct sketching can be adopted through a traditional target detection algorithm.
Further, step 3 optimizes the length and width of the large frame based on the small frame size:
wherein,,
to optimize the length of the rear large frame +.>
In order to optimize the width of the large frame, a is the length of the small frame, b is the width of the small frame, l is the initial length of the large frame image, d is the initial width of the large frame image, int is a rounding function, when decimal points exist, the part after the decimal points is directly removed,
and then the original large frame is replaced by the large frame based on the length and the width of the large frame after optimization, and the defect part is still contained in the new large frame.
Here, it should be noted that, since the sizes of the defects are different, the large rectangular frame is a rectangular frame containing the defects, and the sizes of the rectangular frames are also different. Assuming that the size of the large frame detected by the model is 220×340, the large frame needs to be expanded into a large frame which can be divided by the size of the small frame, the size of the small frame is 97×97, the height 220/97=2.3 of the large frame is approximately 3, the width 340/97 of the large frame is approximately 4, and the number of unclassified small frames is 3×4=12. The expansion can keep the coordinates unchanged at any angle of the initial large frame to optimize
And->
And drawing the optimized large frame as the length and the width. The lower right corner of the small frame is used as the expansion starting point, the expansion direction is up, down, left and right, when the number above the expansion starting point is not enough to expand, the part above the expansion starting point is expanded downwards, and when the number on the left of the expansion starting point is not enough to expand, the part on the left is expanded on the right.
Preferably, the step 4 rectangular box size is 97×97 pixels. Here, the length of the small frame needs to be fixed because the picture is collected by the collection vehicle, the viewing angle and the focal length are fixed, and the picture is connected with the actual road by the camera coordinates. When the length of the small frame is fixed, the area of the road defect in reality can be obtained by the collection of the defect small frames on the image through three-dimensional coordinate transformation, the follow-up fine management of the road maintenance is more convenient, and the area of the road defect is solved based on the defect image, wherein the specific formula is as follows:
wherein S is the defect area, a is the length of the small frame, b is the width of the small frame, n is the total number of the small frames with defects, and alpha is the conversion coefficient of the image length and the actual length.
The bridge connection algorithm is the most core design point of the invention, as shown in fig. 4, the defects of the same type, which are judged by the distance threshold value and are very close, are separated, but based on experience, one defect of the road is continuous, which is not in line with the rule, and even if the condition of missing detection exists at the moment, the bridge connection algorithm is designed for optimization, and the step 5 comprises the following steps:
step 5.1: for any large frame image, finding out all the defect small frames which are classified into the two categories and are judged to be defects;
step 5.2: collecting a plurality of defect small frames of the same category adjacent to each other to construct an island;
step 5.3: respectively calculating the shortest distance between two islands;
step 5.4: and when the shortest distance is smaller than or equal to the distance threshold, bridging the two islands, and updating the two classification categories of the connected small frames into defects, wherein the threshold of bridging is three times the frame length of the small frames.
Preferably, the two island bridge connection comprises the following steps:
step 5.4.1: firstly, for two islands with the shortest distance less than or equal to a distance threshold D, finding out end points L at two ends of the shortest distance 1 And L 2 The coordinates are respectively denoted as (x) 1 ,y 1 ) And (x) 2 ,y 2 ）；
Drawn through L 1 And L 2 Is expressed as:
wherein x and y are the abscissa and ordinate of any point on the line, respectively;
finding out two classification non-defective small frames, and calculating the distance d from the non-defective small frames m to the straight line m :
Wherein d m Representing the distance from the non-defective small frame m to a straight line, x m And y m Respectively representing the abscissa and the ordinate of the non-defective small frame m, wherein A, B and C are calculated using the following formulas, respectively:
step 5.4.2: distance d from non-defective small frame m to straight line m And when the class adjustment threshold value d is less than or equal to, the non-defect small frame is used as a bridge, and the two classes are adjusted to be defects.
The class and position of the large frame are directly derived compared to commonly used multi-class target detection algorithms. The invention can improve the detection precision of road defects. A complex task is divided into a plurality of simple tasks. Model training is easier and the parameters of the two models are not shared. The method has the advantages that the method can optimize the classification detection model without influencing the classification model, and improves the detection precision to a certain extent.
The present invention provides a three-segment road defect image processing method, which is only for illustrating the technical concept and features of the present invention, and is intended to enable those skilled in the art to understand the content of the present invention and implement the same, and is not intended to limit the scope of the present invention. All equivalent changes or modifications made in accordance with the spirit of the present invention should be construed to be included in the scope of the present invention.
Claims (10)
1. The three-section road defect image processing method is characterized by comprising the following steps of:
step 1: collecting road images with defects;
step 2: detecting the position and defect type of a road defect in a road image;
step 3: for any defect, cutting through a rectangular large frame to generate a large frame image containing a type of defect;
step 4: cutting any large frame image into a plurality of rectangular small frames, and judging whether the two classes of the small frames are defects or non-defects;
step 5: judging whether the two classification categories of the small frame are defects or not by optimizing and adjusting based on a bridge connection algorithm;
step 6: the collection of all small frames in the large frame image, which are judged to be defective by all the two classification categories, is recorded as a defect image.
2. The three-stage road defect image processing method according to claim 1, characterized in that:
step 5 comprises the following steps:
step 5.1: for any large frame image, finding out all the defect small frames which are classified into the two categories and are judged to be defects;
step 5.2: collecting a plurality of defect small frames of the same category adjacent to each other to construct an island;
step 5.3: respectively calculating the shortest distance between two islands;
step 5.4: and when the shortest distance is smaller than or equal to the distance threshold value, bridging the two islands, and updating the two classification categories of the connected small frames into defects.
3. The three-segment road defect image processing method according to claim 2, characterized in that:
the connection of the two island bridges comprises the following steps:
step 5.4.1: head partFirstly, for two islands with the shortest distance less than or equal to a distance threshold D, finding out end points L at two ends of the shortest distance 1 And L 2 The coordinates are respectively denoted as (x) 1 ,y 1 ) And (x) 2 ,y 2 ）；
Drawn through L 1 And L 2 Is expressed as:
wherein x and y are the abscissa and ordinate of any point on the line, respectively;
finding out two classification non-defective small frames, and calculating the distance d from the non-defective small frames m to the straight line m :
Wherein d m Representing the distance from the non-defective small frame m to a straight line, x m And y m Respectively representing the abscissa and the ordinate of the non-defective small frame m, wherein A, B and C are calculated using the following formulas, respectively:
step 5.4.2: distance d from non-defective small frame m to straight line m And when the class adjustment threshold value d is less than or equal to, the non-defect small frame is used as a bridge, and the two classes are adjusted to be defects.
4. The three-stage road defect image processing method according to claim 1, characterized in that:
step 3, optimizing the length and width of the large frame based on the size of the small frame:
wherein,,
to optimize the length of the rear large frame +.>
In order to optimize the width of the large frame, a is the length of the small frame, b is the width of the small frame, l is the initial length of the large frame image, d is the initial width of the large frame image, int is a rounding function, when decimal points exist, the part after the decimal points is directly removed,
and then the original large frame is replaced by the large frame based on the length and the width of the large frame after optimization, and the defect part is still contained in the new large frame.
5. The three-segment road defect image processing method according to claim 4, wherein:
keeping the coordinates unchanged at any angle of the initial large frame to optimize
And->
And drawing the optimized large frame as the length and the width.
6. The three-stage road defect image processing method according to claim 1, characterized in that:
solving the area of the road defect based on the defect image, wherein the specific formula is as follows:
wherein S is the defect area, a is the length of the small frame, b is the width of the small frame, n is the total number of the small frames with defects, and alpha is the conversion coefficient of the image length and the actual length.
7. The three-stage road defect image processing method according to claim 1, characterized in that:
and 3, carrying out corrosion and expansion treatment on the large frame image to remove noise in the image, so that road defects in the large frame and the background edge are more easily distinguished.
8. The three-stage road defect image processing method according to claim 1, characterized in that:
and 2, detecting the defect type and the defect position based on the deep learning target detection model.
9. The three-stage road defect image processing method according to claim 1, characterized in that:
and 3, drawing a large frame image through a rectangular frame.
10. The three-stage road defect image processing method according to claim 1, characterized in that:
step 4 rectangular small box size is 97 x 97 pixels.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310677633.5A CN116433659B (en) | 2023-06-09 | 2023-06-09 | Three-section road defect image processing method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310677633.5A CN116433659B (en) | 2023-06-09 | 2023-06-09 | Three-section road defect image processing method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116433659A true CN116433659A (en) | 2023-07-14 |
| CN116433659B CN116433659B (en) | 2023-08-29 |
Family
ID=87089375
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310677633.5A Active CN116433659B (en) | 2023-06-09 | 2023-06-09 | Three-section road defect image processing method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116433659B (en) |
Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH09307217A (en) * | 1996-05-13 | 1997-11-28 | Ntn Corp | Defect correcting method and device of continuous pattern |
| WO2018000731A1 (en) * | 2016-06-28 | 2018-01-04 | 华南理工大学 | Method for automatically detecting curved surface defect and device thereof |
| US20200160083A1 (en) * | 2018-11-15 | 2020-05-21 | International Business Machines Corporation | Efficient defect localization/segmentation for surface defect inspection |
| CN111402248A (en) * | 2020-03-23 | 2020-07-10 | 华南理工大学 | Transmission line lead defect detection method based on machine vision |
| CN111695482A (en) * | 2020-06-04 | 2020-09-22 | 华油钢管有限公司 | Pipeline defect identification method |
| CN112083017A (en) * | 2020-09-10 | 2020-12-15 | 上海航天精密机械研究所 | Intelligent detection device, method and medium for internal defects of welding seam |
| CN113160184A (en) * | 2021-04-26 | 2021-07-23 | 贵州电网有限责任公司 | Unmanned aerial vehicle intelligent inspection cable surface defect detection method based on deep learning |
| CN113160192A (en) * | 2021-04-28 | 2021-07-23 | 北京科技大学 | Visual sense-based snow pressing vehicle appearance defect detection method and device under complex background |
| US20210319265A1 (en) * | 2020-11-02 | 2021-10-14 | Zhengzhou University | Method for segmentation of underground drainage pipeline defects based on full convolutional neural network |
| US20210374940A1 (en) * | 2019-12-30 | 2021-12-02 | Goertek Inc. | Product defect detection method, device and system |
| CN113870263A (en) * | 2021-12-02 | 2021-12-31 | 湖南大学 | Real-time monitoring method and system for pavement defect damage |
| WO2022036919A1 (en) * | 2020-08-17 | 2022-02-24 | 深圳市商汤科技有限公司 | Defect detection method and apparatus, and electronic device and computer storage medium |
| KR102396169B1 (en) * | 2021-09-07 | 2022-05-11 | 주식회사 로드텍 | Apparatus and method for evaluating state of road pavement |
| CN115511784A (en) * | 2022-08-17 | 2022-12-23 | 浙江理工大学 | Intelligent road defect identification method based on deep convolution generated countermeasure network |
-
2023
- 2023-06-09 CN CN202310677633.5A patent/CN116433659B/en active Active
Patent Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH09307217A (en) * | 1996-05-13 | 1997-11-28 | Ntn Corp | Defect correcting method and device of continuous pattern |
| WO2018000731A1 (en) * | 2016-06-28 | 2018-01-04 | 华南理工大学 | Method for automatically detecting curved surface defect and device thereof |
| US20200160083A1 (en) * | 2018-11-15 | 2020-05-21 | International Business Machines Corporation | Efficient defect localization/segmentation for surface defect inspection |
| US20210374940A1 (en) * | 2019-12-30 | 2021-12-02 | Goertek Inc. | Product defect detection method, device and system |
| CN111402248A (en) * | 2020-03-23 | 2020-07-10 | 华南理工大学 | Transmission line lead defect detection method based on machine vision |
| CN111695482A (en) * | 2020-06-04 | 2020-09-22 | 华油钢管有限公司 | Pipeline defect identification method |
| WO2022036919A1 (en) * | 2020-08-17 | 2022-02-24 | 深圳市商汤科技有限公司 | Defect detection method and apparatus, and electronic device and computer storage medium |
| CN112083017A (en) * | 2020-09-10 | 2020-12-15 | 上海航天精密机械研究所 | Intelligent detection device, method and medium for internal defects of welding seam |
| WO2022053001A1 (en) * | 2020-09-10 | 2022-03-17 | 上海航天精密机械研究所 | Weld seam internal defect intelligent detection device and method, and medium |
| US20210319265A1 (en) * | 2020-11-02 | 2021-10-14 | Zhengzhou University | Method for segmentation of underground drainage pipeline defects based on full convolutional neural network |
| CN113160184A (en) * | 2021-04-26 | 2021-07-23 | 贵州电网有限责任公司 | Unmanned aerial vehicle intelligent inspection cable surface defect detection method based on deep learning |
| CN113160192A (en) * | 2021-04-28 | 2021-07-23 | 北京科技大学 | Visual sense-based snow pressing vehicle appearance defect detection method and device under complex background |
| KR102396169B1 (en) * | 2021-09-07 | 2022-05-11 | 주식회사 로드텍 | Apparatus and method for evaluating state of road pavement |
| CN113870263A (en) * | 2021-12-02 | 2021-12-31 | 湖南大学 | Real-time monitoring method and system for pavement defect damage |
| CN115511784A (en) * | 2022-08-17 | 2022-12-23 | 浙江理工大学 | Intelligent road defect identification method based on deep convolution generated countermeasure network |
Non-Patent Citations (3)
| Title |
|---|
| 官鑫;赵智雅;高晓蓉;: "图像处理技术在钢轨表面缺陷检测和分类中的应用", 铁路计算机应用, no. 06 * |
| 常海涛;苟军年;李晓梅;: "Faster R-CNN在工业CT图像缺陷检测中的应用", 中国图象图形学报, no. 07 * |
| 郭慧;王霄;刘传泽;周玉成;: "基于灰度共生矩阵和分层聚类的刨花板表面图像缺陷提取方法", 林业科学, no. 11 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116433659B (en) | 2023-08-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107316064B (en) | Asphalt pavement crack classification and identification method based on convolutional neural network | |
| US10489903B2 (en) | Stepwise refinement detection method for pavement cracks | |
| CN104021574B (en) | Pavement disease automatic identifying method | |
| CN112001411B (en) | Dam crack detection algorithm based on FPN structure | |
| CN113469951B (en) | Hub defect detection method based on cascade region convolutional neural network | |
| CN111798447B (en) | Deep learning plasticized material defect detection method based on fast RCNN | |
| CN113409267B (en) | Pavement crack detection and segmentation method based on deep learning | |
| CN110956207B (en) | Method for detecting full-element change of optical remote sensing image | |
| CN113327255A (en) | Power transmission line inspection image processing method based on YOLOv3 detection, positioning and cutting and fine-tune | |
| CN111079822A (en) | Method for identifying dislocation fault image of middle rubber and upper and lower plates of axle box rubber pad | |
| CN116109616A (en) | Pavement crack detection and small-surface element fitting detection method based on YOLOv5 | |
| CN115294541A (en) | Local feature enhanced Transformer road crack detection method | |
| CN116503336A (en) | Pavement crack detection method based on deep learning | |
| CN115205255A (en) | Stone automatic grading method and system based on deep learning | |
| Zhao et al. | A method of crack detection based on convolutional neural networks | |
| CN116433659B (en) | Three-section road defect image processing method | |
| CN113762247A (en) | Road crack automatic detection method based on significant instance segmentation algorithm | |
| CN107992799A (en) | Towards the preprocess method of Smoke Detection application | |
| CN116958837A (en) | Municipal facilities fault detection system based on unmanned aerial vehicle | |
| Gao et al. | Intelligent crack damage detection system in shield tunnel using combination of retinanet and optimal adaptive selection | |
| CN116503359A (en) | Road surface crack monitoring method based on road side camera | |
| CN115578725B (en) | Landslide boundary and crack marking method and system based on man-machine cooperation | |
| CN116363426A (en) | Automatic detection method for health state of highway antiglare shield | |
| Tian et al. | Road crack detection algorithm based on YOLOv3 | |
| CN115511784A (en) | Intelligent road defect identification method based on deep convolution generated countermeasure network |
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 |