CN115457041A - Road quality identification and detection method - Google Patents
Road quality identification and detection method Download PDFInfo
- Publication number
- CN115457041A CN115457041A CN202211417318.0A CN202211417318A CN115457041A CN 115457041 A CN115457041 A CN 115457041A CN 202211417318 A CN202211417318 A CN 202211417318A CN 115457041 A CN115457041 A CN 115457041A
- Authority
- CN
- China
- Prior art keywords
- gray
- defect
- pixel
- window
- value
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 27
- 230000007547 defect Effects 0.000 claims abstract description 248
- 230000011218 segmentation Effects 0.000 claims abstract description 61
- 239000011159 matrix material Substances 0.000 claims abstract description 45
- 238000000034 method Methods 0.000 claims description 39
- 238000005065 mining Methods 0.000 claims 1
- 230000002950 deficient Effects 0.000 description 6
- 230000000694 effects Effects 0.000 description 6
- 238000010606 normalization Methods 0.000 description 6
- 230000008569 process Effects 0.000 description 6
- 238000004458 analytical method Methods 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 4
- 241001270131 Agaricus moelleri Species 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 2
- 238000003709 image segmentation Methods 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000006467 substitution reaction Methods 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/40—Image enhancement or restoration by the use of histogram techniques
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- 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
- G06T2207/30248—Vehicle exterior or interior
- G06T2207/30252—Vehicle exterior; Vicinity of vehicle
- G06T2207/30256—Lane; Road marking
Abstract
The invention relates to the technical field of image processing, in particular to a road quality identification and detection method, which comprises the following steps: acquiring a road surface gray image, and segmenting the road surface gray image to acquire a lane line area so as to obtain a preferred road image; processing the preferred road image by using a window with a set size, calculating the gray level uniformity degree corresponding to the window according to the gray level of a pixel point in the window, and determining a possible defect area; detecting a voting value according to Hough lines of pixel points in a possible defect area after skeleton refinement to obtain the defect probability of the pixel points, and further obtaining the defect probability value corresponding to the gray level; and constructing a defect characteristic matrix according to the defect probability value, determining the size of the superpixel block according to the energy characteristic value of the matrix, further performing superpixel segmentation to obtain a defect region, and then performing identification to obtain a quality identification result. The invention can improve the segmentation efficiency and ensure the segmentation precision.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a road quality identification and detection method.
Background
In recent years, road networks are developed more and more in China, the number of roads is increased more and more, damages to the roads caused by vehicle use are increased, and depressions, cracks and the like in the roads have great threats to vehicle and personnel safety, so that attention needs to be paid to road quality detection. The unmanned aerial vehicle road inspection system is developed gradually, an unmanned aerial vehicle is mainly used for collecting images, and then whether defects exist in the images is analyzed, so that whether the quality of a current road section is qualified or not is judged.
The super-pixel segmentation effect in the existing image segmentation technology is good, but the super-pixel segmentation needs to set more parameters which are globally fixed and unchangeable in the conventional situation, the road surface situation may be complex and various, and one parameter cannot accurately segment the defect that the road surface has the complex situation.
Disclosure of Invention
In order to solve the above technical problems, an object of the present invention is to provide a road quality identification and detection method, which adopts the following technical solutions:
acquiring a road surface gray image, segmenting the road surface gray image to acquire a lane line region, and assigning a pixel value of a pixel point in the lane line region by using a gray value with the highest frequency in the road surface gray image to obtain an optimal road image;
dividing the pixel values of the pixels in the preferred road image into gray levels, processing the preferred road image by using a window with a set size, and calculating the gray uniformity degree corresponding to the window according to the gray levels of the pixels in the window; determining a possible defect area according to the gray level uniformity degree corresponding to the window;
skeleton refinement is carried out on each possible defect area, hough line detection is carried out on each possible defect area after skeleton refinement, and the defect possibility of pixel points is obtained according to Hough line detection voting values of the pixel points in the possible defect areas; obtaining a defect probability value corresponding to the gray level according to the mean value of the defect possibilities of the pixel points belonging to the same gray level;
constructing a defect feature matrix according to defect probability values corresponding to the gray levels of the pixel points, obtaining energy feature values according to the defect probability values in the defect feature matrix, and determining the size of the super-pixel block according to the energy feature values; dividing the possible defect area by using a superpixel division method according to the size of the superpixel block to obtain a defect area; and determining the defect type according to the defect area, and further identifying the road quality to obtain a quality identification result.
Preferably, the segmenting the road surface grayscale image to obtain the lane line region specifically includes:
and (3) counting gray values of pixel points in the gray image on the road surface to construct a gray histogram, determining a gray range according to the gray histogram, and selecting the pixel points with the gray values in the gray range as initial seed points to perform region growth to obtain a lane line region.
Preferably, the dividing of the gray scale of the pixel value of the pixel point in the preferred road image specifically includes:
and counting gray values of pixel points in the optimized road image to construct a gray histogram, segmenting the gray histogram by using a multi-threshold segmentation method, recording the gray values segmented in the same interval as the same gray level, sequencing the gray levels according to the sequence of the gray values from small to large, and recording the sequence numbers as the corresponding gray levels according to the sequence.
Preferably, the method for obtaining the gray level uniformity degree corresponding to the window specifically comprises:
wherein HD represents the gray level uniformity degree corresponding to the window,
represents the gray scale of the a-th pixel point,
representing the total number of pixels in the window, exp () represents an exponential function with a natural constant e as the base.
Preferably, the determining the possible defect area according to the gray uniformity degree corresponding to the window specifically includes:
recording the set size of the window as an initial size, carrying out mine enlargement on the initial size of the window according to the set step length when the gray level uniformity degree corresponding to the window of the initial size is smaller than or equal to a degree threshold, calculating the gray level uniformity degree corresponding to the enlarged window, continuing to enlarge the size of the window when the gray level uniformity degree corresponding to the enlarged window is still smaller than or equal to the degree threshold, and recording the area in the window as a possible defect area when the gray level uniformity degree corresponding to the enlarged window is larger than the degree threshold.
Preferably, the constructing of the defect feature matrix according to the defect probability value corresponding to the gray level of the pixel point specifically comprises:
and constructing a gray level co-occurrence matrix corresponding to a possible defect area according to the gray level of the pixel point, and constructing a defect feature matrix according to the defect probability value corresponding to the gray level of the pixel point and elements in the gray level co-occurrence matrix.
Preferably, the determining the size of the super-pixel block according to the energy characteristic value specifically includes:
wherein S is the size of the super pixel block, ma represents the energy characteristic value corresponding to the defect characteristic matrix,
represents a constant coefficient, e is a natural constant,
as a function of the rounding-up.
Preferably, the dividing the possible defect region according to the size of the super pixel block by using the super pixel dividing method specifically includes:
determining a search range during superpixel segmentation according to the size of the superpixel block; determining segmentation seed points according to gradient values of pixel points in a possible defect region, calculating an inverse difference moment of a defect characteristic matrix, and determining a weighted value corresponding to a distance according to the inverse difference moment; calculating the distance between the pixel point and the segmentation seed point in the corresponding search range according to the weight value, the gray value of the pixel point in the possible defect region and the gray value of the segmentation seed point, and the pixel coordinates of the pixel point in the possible defect region and the segmentation seed point; and performing super-pixel segmentation according to the distance.
The embodiment of the invention at least has the following beneficial effects:
the method comprises the steps of obtaining a lane line area by segmenting a road surface gray level image, further obtaining an optimal road image, considering the influence of a lane line part existing on the road surface on a defect part, and removing the part belonging to the lane line, so that the subsequent defect segmentation is not interfered by the lane line part; calculating the gray level uniformity degree in the surplus in the optimized road image, further determining possible defect areas, carrying out block analysis on the image, roughly judging the areas possibly having defects through the distribution uniformity of the gray level values of the pixels in a single window, and further reducing the range for carrying out defect segmentation; the defect probability of the pixel points is obtained according to the Hough line detection voting value of the pixel points in the possible defect area, the defect probability value corresponding to the gray level is further obtained, the shape characteristics of the defects are considered, the pixel points are analyzed to obtain the possible degree that the pixel points are the defects, the defect parts and the dirty parts can be distinguished, and the precision of subsequent super-pixel segmentation is further ensured; the size of the superpixel block is determined according to the energy characteristic value, the possible defect area is subjected to superpixel segmentation to obtain the defect area, and the quality identification result of the road is obtained.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions and advantages of the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
Fig. 1 is a flow chart of a method of the road quality identification detection method of the present invention.
Detailed Description
To further illustrate the technical means and effects of the present invention adopted to achieve the predetermined objects, the following detailed description will be given to a road quality identification and detection method according to the present invention, with reference to the accompanying drawings and preferred embodiments, and the specific implementation, structure, features and effects thereof. In the following description, different "one embodiment" or "another embodiment" refers to not necessarily the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
The following describes a specific scheme of the road quality identification and detection method provided by the invention in detail with reference to the accompanying drawings.
Example (b):
referring to fig. 1, a flowchart of a method for identifying and detecting road quality according to an embodiment of the present invention is shown, where the method includes the following steps:
the method comprises the steps of firstly, obtaining a road surface gray image, segmenting the road surface gray image to obtain a lane line area, and assigning values to pixel values of pixels in the lane line area by utilizing the gray value with the highest frequency of occurrence in the road surface gray image to obtain an optimal road image.
Firstly, in this embodiment, through setting up unmanned aerial vehicle's flight route and flying height to control unmanned aerial vehicle during the flight and gather clear road surface image, because the collection process may receive mechanical noise's influence, the image transmission's process may receive impulse noise's influence, consequently carries out noise reduction to the image of gathering, adopts gaussian filter to carry out noise reduction to road surface image in this embodiment. And performing semantic segmentation on the filtered image, removing the interference of a background area outside the road, and only keeping the surface part of the road. And meanwhile, carrying out graying processing on the image without the background to obtain a road surface grayscale image. The specific steps of semantic segmentation are well-known technologies, and are not described herein too much, and an implementer may also select another suitable method to remove the interference of the background according to the actual situation.
It should be noted that there is generally a white lane line on the road surface, or the longer the service life of the road is, the more the lane line on the road is worn, the smaller the gray value corresponding to the white lane line becomes, when detecting whether there is a damage on the road surface, the interference of the lane line on the texture of the image is larger, and the lane line does not belong to the texture of the damaged defect portion, so the lane line portion needs to be removed.
Then, the road surface gray image is segmented to obtain a lane line region, specifically, the gray values of pixel points in the road surface gray image are counted to construct a gray histogram, a gray range is determined according to the gray histogram, and the pixel points with the gray values within the gray range are selected as initial seed points to perform region growth to obtain the lane line region.
On the road surface gray image, gray values of pixel points belonging to the lane line part are large, so that the gray values of the pixel points in the road surface gray image are counted to construct a gray histogram, curve fitting is carried out on the gray histogram to obtain a gray curve, and peaks and troughs on the gray curve are obtained. And selecting a trough adjacent to the peak with the maximum gray value as a segmentation point, wherein the gray value corresponding to the trough is smaller than the gray value corresponding to the peak with the maximum gray value, namely the trough is positioned on the left side of the peak with the maximum gray value.
Because the gray value of the pixel point belonging to the lane line part is larger, the gray value corresponding to the peak with the largest gray value in the gray curve is probably the gray value corresponding to most pixel points in the lane line part, so the nearest wave trough on the left side of the peak with the largest gray value is selected as the segmentation point, and the approximate range of the gray value belonging to the lane line part can be determined. That is, all gray values on the right side of the segmentation point form a gray range, and pixel points of the gray value in the gray range on the road surface gray image are all possible pixel points of the lane line part.
Therefore, any pixel point with the gray value within the gray range on the road surface gray image is selected, the pixel point is used as an initial seed point to carry out region growing to obtain a lane line region, and a rule implementer who carries out the region growing by using the initial seed point can set the pixel point according to actual conditions. And after the current region growth is finished, continuously selecting initial growth points from the pixel points with the residual gray values within the gray range to perform region growth, and continuously performing region growth until the pixel points with the gray values within the gray range are traversed once, so that the growth of all lane line regions is finished.
Finally, after the lane line area is obtained, the lane line portion needs to be removed because the lane line portion has certain regularity and the texture occupying a relatively large area in the road surface has certain interference to the defective texture existing on the road surface. Specifically, a gray value with the highest frequency of occurrence of the road surface gray image value is obtained, that is, a gray value corresponding to a peak with the highest peak value in a gray curve is obtained, and the gray value is used for re-assigning the pixel value of the pixel point in the lane line region to obtain the preferred road image. Because the most main color in the road surface gray image is the color of the road main body part, the gray value with the highest frequency of occurrence must belong to the road surface main body part, the gray value of the lane line part is replaced by the gray value of the road main body part, the interference of the lane line can be removed, and the defect detection result of the road main body part cannot be influenced by new factors.
Dividing gray scales of pixel values of pixels in the preferred road image, processing the preferred road image by using a window with a set size, and calculating the gray uniformity degree corresponding to the window according to the gray scales of the pixels in the window; and determining a possible defect area according to the gray level uniformity degree corresponding to the window.
First, it should be noted that, since all the preferred road images include the road surface portion, if there is no damage defect on the road surface, there is no more complex texture in the preferred road images, and if there is a damage defect on the road surface, after removing the lane line portion, the only texture in the preferred road images is the region texture with the damage defect. Therefore, the preferred road image needs to be processed in a blocking mode, whether defects exist in the corresponding area is judged according to the texture information of each block, and rough defect information is obtained so that the image can be accurately segmented subsequently.
Specifically, the pixel values of the pixels in the preferred road image are classified into gray levels, in this embodiment, the gray levels of the pixels in the preferred road image are counted to construct a gray histogram, the gray histogram is segmented by using a multi-threshold segmentation method, the gray levels segmented in the same interval are recorded as the same gray level, the gray levels are sorted according to the sequence of the gray levels from small to large, and the arrangement sequence numbers are recorded as the corresponding gray levels according to the arrangement sequence.
The reason why the multi-threshold segmentation method is used for dividing the gray levels is that in the traditional gray level segmentation method, the gray levels with larger differences may be divided into the same gray level, so that when the divided gray levels are used for analyzing an image, a defect part and a road surface part may be divided into the same gray level, and further, the image distortion condition is caused, and the image segmentation precision is influenced. When the Zengjin threshold segmentation method is used for multi-threshold segmentation, the gray values with larger differences can be divided into different gray levels as much as possible, meanwhile, the levels are divided according to the gray values of all pixel points in the preferred road image, so that the method is more suitable for the image, and the gray difference between the defective part and the normal part in the image can be increased as much as possible.
Then, analyzing the preferred road image according to the gray scale can reduce the amount of calculation. The preferred road image is processed by using the window with the set size, in this embodiment, the window with the set size of 3 × 3, that is, the window with the size of 3 × 3 slides from the position of the upper left corner of the preferred road image to the left to the right and from the top to the bottom. The purpose of processing the preferred road image by using the window is to perform block analysis on the image and acquire a possibly defective area.
When the area in the window is analyzed, if the area in the window has no damage defect, the area in the window is the road surface part, the gray information in the window is uniform, namely the difference between the gray level of each pixel point in the window and the gray level of the whole area in the window is small. If the area in the window has damage defect, the area in the window has defect part except the road surface part, so the gray information in the window is more disordered, and the difference between the gray level of each pixel point in the window and the gray level of the whole area in the window is larger.
Based on the method, the gray level uniformity degree corresponding to the window is calculated according to the gray level of the pixel points in the window, and the gray level uniformity degree is expressed by a formula as follows:
wherein HD represents the degree of gray level uniformity corresponding to the window,
represents the gray scale of the a-th pixel point,
representing the total number of pixels in the window, exp () represents an exponential function with a natural constant e as the base.
And finally, determining a possible defect area according to the size of the window and the gray level uniformity degree corresponding to the window, recording the set size of the window as an initial size, carrying out mine enlargement on the initial size of the window according to a set step length when the gray level uniformity degree corresponding to the window of the initial size is less than or equal to a degree threshold value, calculating the gray level uniformity degree corresponding to the enlarged window, continuing to enlarge the size of the window when the gray level uniformity degree corresponding to the enlarged window is still less than or equal to the degree threshold value, and recording the area in the window as the possible defect area until the gray level uniformity degree corresponding to the enlarged window is greater than the degree threshold value.
In the present embodiment, the degree threshold valueThe value of (b) is 0.4, and the implementer can set the value according to actual conditions. When the degree of uniformity of the gray level corresponding to the window of the initial size is less than or equal to the degree threshold value, that is
When the window is used, it is described that gray information of pixels in the window is not uniform, damage defects may exist in the window, in order to obtain a local maximum defect portion, the size of the window needs to be enlarged, and the initial size of the window is enlarged according to a set step length, that is, the initial size is enlarged from the size of 3 × 3 to the size of 5 × 5. In this embodiment, the step length is set to 2, and the implementer can set the step length according to the actual situation.
Calculating the gray level uniformity degree corresponding to the expanded window, when the gray level uniformity degree corresponding to the expanded window is smaller than or equal to a degree threshold, continuing to expand the size of the window according to a set step length until the gray level uniformity degree corresponding to the expanded window is larger than the degree threshold, indicating that the gray level information uniformity of the pixel points in the window is relatively large at the moment, indicating that the window contains more pixel points belonging to the surface part of the road and the window contains a local maximum defect part, and recording the window as a region possibly having defects for accurately segmenting out the region with the damage defects in the follow-up process, not performing block processing on the other normal parts, and analyzing the region possibly having the damage defects in the follow-up process.
Step three, skeleton refinement is carried out on each possible defect area, hough line detection is carried out on each possible defect area after skeleton refinement, and the defect possibility of the pixel points is obtained according to Hough line detection voting values of the pixel points in the possible defect areas; and obtaining the defect probability value corresponding to the gray level according to the mean value of the defect probability of the pixel points belonging to the same gray level.
First, it should be noted that the possible damage on the road surface generally includes crack defects, and the texture information of the road crack damage defect represented in the preferred road image is relatively disordered. Due to the fact that the service life of the road is long, dirty parts such as oil stains and the like can exist on the surface of the road besides crack damage defects, and the existence of the dirty parts can influence the follow-up segmentation of the defect area, the dirty parts and the defect parts need to be distinguished firstly. In the implementation, each possible defect area is analyzed, and the possibility that each pixel point in the possible defect area is a lost defect part pixel point is obtained.
The shape of the connected domain of the texture information part contained in the possible defect region may be various, wherein the more regular part is a part where a crack defect exists, and the less regular part is a dirty part. Meanwhile, the part with the crack defect has a relatively intuitive straight line characteristic, and the shape change of the dirty part has no regularity and does not accord with the straight line characteristic. Based on this feature, it is possible to distinguish between a damaged defective portion and a dirty portion within a possible defective region.
Then, skeleton refinement is performed on each possible defect region, and in this embodiment, when the linearity of the possible defect region is analyzed without skeleton refinement, the distinguishing effect is not obvious because a plurality of line segments may exist in all the regions. And after the possible defect regions are subjected to skeleton thinning, only the central axis of each region is obtained, the central axis is a line with single line width, the crack defect portion is thinned to form a straight line segment, the dirty portion is thinned to form a coil, and the crack defect portion and the dirty portion can be accurately distinguished.
Performing Hough line detection on each possible defect area after skeleton refinement, and obtaining the defect possibility of a pixel point according to the Hough line detection voting value of the pixel point in the possible defect area, namely summing the Hough line detection voting values of the pixel point in the possible defect area, and taking the summed result as the defect possibility of the pixel point, wherein the Hough line detection voting value can reflect whether the pixel point belongs to a straight line, the larger the voting value is, the more likely the pixel point belongs to the straight line, the greater the defect possibility corresponding to the pixel point is, and the more likely the pixel point belongs to a crack defect part.
The defect probability of the pixel point represents the probability that the pixel point is a crack defect and the defect probability of the pixel pointThe larger the value is, the more likely the pixel point belongs to the crack defect part. And normalizing the defect probability of the pixel point, in this embodiment, the normalization process is expressed by the following formula:
wherein M is the defect probability of the pixel points before normalization,
and e is a natural constant, which is the defect probability of the normalized pixel points.
Finally, since each pixel corresponds to a gray scale, there may be a possibility that the gray scales of a plurality of pixels are the same, that is, the same gray scale corresponds to a plurality of pixels, and each pixel corresponds to a defect possibility after normalization, so it is necessary to obtain the defect possibility corresponding to the same gray scale. The method includes the steps that defect probability values corresponding to gray levels are obtained according to the mean values of the defect possibilities of pixel points belonging to the same gray level, specifically, the defect possibilities of the pixel points belonging to the same gray level are summed and averaged, the mean value is used as the defect probability value corresponding to the gray level, each gray level is processed according to the same method, and the only defect probability value corresponding to each gray level can be obtained, namely, each pixel point corresponds to one gray level and also corresponds to one defect probability value.
Fourthly, constructing a defect characteristic matrix according to the defect probability value corresponding to the gray level of the pixel point, obtaining an energy characteristic value according to the defect probability value in the defect characteristic matrix, and determining the size of the superpixel block according to the energy characteristic value; performing segmentation processing on the possible defect area by using a superpixel segmentation method according to the size of the superpixel block to obtain a defect area; and determining the defect type according to the defect area, and further identifying the road quality to obtain a quality identification result.
Firstly, it should be noted that, in the above steps, the possibility that each pixel point may be a crack defect is obtained, and then texture information in a possible defect region needs to be analyzed according to the possibility, if the number of pixel points belonging to a crack defect portion in the possible defect region is large, the texture in the possible defect region is relatively disordered, and different superpixel sizes need to be adopted when the superpixel segmentation is subsequently performed on the possible defect region according to different texture information, so that an optimal segmentation effect is achieved.
In this embodiment, a defect feature matrix corresponding to a possible defect region is obtained by using an analogy method for obtaining a gray level co-occurrence matrix, because the size of the obtained super pixel needs to be matched with the thickness of a crack portion, when the texture of the possible defect region is analyzed by using the gray level co-occurrence matrix, only the texture distribution condition of the whole possible defect region is analyzed, the crack defect and the dirty portion contained in the possible defect region may not be distinguished, the selection of the size of the super pixel is affected, and further the accuracy of super pixel segmentation is affected.
Specifically, a gray level co-occurrence matrix of a possible defect area is obtained, the gray level co-occurrence matrix is updated by using the defect probability value corresponding to the gray level of the pixel point, and a defect feature matrix is constructed. In this embodiment, the number of times of occurrence of each gray scale pair in the directions of 0 °, 45 °, 90 °, and 135 ° is counted, and a gray scale co-occurrence matrix in 4 directions is obtained in sequence for each possible defect region. Acquiring the gray frequency number in the defect characteristic matrix according to the defect probability value corresponding to the gray levels of the elements and the pixels in the gray level co-occurrence matrix, and expressing the gray frequency number as follows by using a formula:
wherein the content of the first and second substances,
representing the frequency of the gray level pairs corresponding to gray level i and gray level j in the defect feature matrix,
indicating the probability value of the defect corresponding to the gray level i,
representing the probability value of the defect corresponding to the gray level j,
and expressing the frequency of the gray level pairs corresponding to the gray level i and the gray level j in the gray level co-occurrence matrix.
The gray level co-occurrence matrix is updated by utilizing the defect probability value, the possibility that each pixel point is possibly a crack defect is combined, so that when texture information in a possible defect area is analyzed, a crack defect part and a dirty part can be distinguished, and the characteristic value of the crack defect part can be more highlighted. Multiplying elements in the gray level co-occurrence matrix by using the defect probability values corresponding to the gray levels, wherein the more probable the pixel points corresponding to the gray levels are crack defects, the larger the value of the corresponding elements in the defect characteristic matrix is.
Then, the mean value of the energy values corresponding to the defect feature matrix in the four directions is calculated as the energy feature value corresponding to the defect feature matrix, and the calculation method of the energy feature value is the same as that of the energy value in the gray level co-occurrence matrix, and is not described herein more. The energy characteristic value reflects the gray level in the possible defect area, the defect distribution uniformity degree and the texture thickness. If the element values of the defect feature matrix are similar, the energy feature value is smaller, and the texture is detailed; if some of the values are large and others are small, the energy characteristic value is large. A large energy eigenvalue indicates a more uniform and regularly varying texture pattern. The energy characteristic value reflects the thickness degree of the texture, the thickness degree of the texture has important reference value for subsequently selecting the size of the superpixel, the thicker texture needs to obtain the larger size of the superpixel, and the thinner texture needs to obtain the smaller size of the superpixel block.
The thicker or thinner texture refers to the thickness of the width of a crack defect part in a possible defect area, and the wider crack needs a larger super-pixel block to be completely segmented without causing under-segmentation; the thinner the crack, the smaller the superpixel block is needed to just segment the crack, and the condition of excessive segmentation cannot be caused. Therefore, the proper size of the superpixel block needs to be adaptively selected in combination with the thickness of the crack, so that the optimal segmentation effect is achieved.
Specifically, the size of the superpixel block is determined according to the energy characteristic value, and is expressed by a formula:
wherein S is the size of the super pixel block, ma represents the energy characteristic value corresponding to the defect characteristic matrix,
represents a constant coefficient, e is a natural constant,
as a function of the rounding-up. The energy characteristic value corresponding to the conventional texture is large in range, so that the energy characteristic value is normalized, and the normalized energy characteristic value is multiplied by a constant coefficient to enable the energy characteristic value to be adaptive to a pixel unit.
It should be noted that, after obtaining the super-pixel size for super-pixel segmentation of the possible defect region, the size of the super-pixel block is compared with the size of the current possible defect region, and if the size of the super-pixel block is larger than the size of the possible defect region, which indicates that the defect portion included in the current possible defect region is larger, the size of the super-pixel block during super-pixel segmentation corresponding to the current possible defect region needs to be set to the size of the possible defect region.
Further, in this embodiment, each possible defect region is divided by using the SLIC super-pixel division method, and the size of the desired super-pixel block is determined to be S × S, and if the size of the initial cluster center required for processing by using the SLIC super-pixel division method is the same as the size of the super-pixel block, the search range is 2s × 2s. The relevant content of the SLIC super-pixel segmentation method is known in the art and will not be described herein too much.
Calculating the local texture correlation of each possible defect area, judging whether the gray distance accounts for a larger weight or the position distance accounts for a larger weight when the super-pixel segmentation is carried out on the pixel points in the current possible defect area according to the local correlation, and endowing a proper weight value for the gray distance and the position distance of the pixel points in each possible defect area. Specifically, the inverse difference matrix of the defect feature matrix corresponding to the possible defect region is calculated, and the calculation method of the inverse difference matrix is the same as the calculation method of the inverse difference matrix of the gray level co-occurrence matrix, and is not described herein too much.
The texture inverse difference moment of each possible defect area reflects the homogeneity of the texture in the current possible defect area and is used for measuring the change of the local texture of the image, when each local texture in the possible defect area does not change, the larger the inverse difference moment value is, the more uniform the local gray scale is, otherwise, if the inverse difference moment value in the possible defect area is smaller, the more non-uniform the gray scale value and the defect probability value in the possible defect area are.
It should be noted that, the inverse difference moment is used to obtain the homogeneity of the defect texture of each possible defect region, because not only there is a gray level difference between the defect and the normal region, but also there is a gray level difference between the defect and the defect, for example, there is a difference in gray level between a crack with a larger width and a crack with a smaller width, it can be reflected whether there is a defect with a gray level difference in the current possible defect region according to the homogeneity of the crack obtained by the inverse difference moment, and when there is a defect with a gray level difference in the possible defect region, it is indicated that there may be a crack defect and a dirty part in the defective region, so by adjusting the weight value of the gray level distance and the position distance, the gray level distance of the region with a gray level difference is larger, and further, the crack defect and the dirty part can be more accurately segmented.
When the gray value and the defect probability value in the possible defect area are more uneven, the texture of the current possible defect area is considered to have crack defects and dirty parts, and the gray distance and the position distance are required to be combined to comprehensively determine which super-pixel block the current pixel point is divided into.
Meanwhile, since the range of the inverse difference moment is large, normalization is required, and in this embodiment, since the smaller the inverse difference moment is, the more uneven the texture in the region is, normalization is formulated as
Wherein IDM represents the inverse difference moment of the possible defect area,
and expressing the optimized characteristic value after the normalization processing, wherein the larger the value is, the smaller the value of the inverse difference moment is, and the more uneven the texture in the region is.
When the preferred characteristic value is greater than the threshold, the weight value to be assigned to the gray scale distance is
In which
Is a gray-scale distance weight value that,
indicating a preferred characteristic value. If the texture in the region is uneven, it indicates that there may be a crack defect portion and a dirty portion in the region, and a weight value corresponding to the gray level distance and the position distance needs to be set according to the uniformity of the texture distribution. The weight value given to the position distance is
Wherein
The position distance weight value is the position distance weight value, when the position distance weight is larger, the probability that the pixel point is distributed to the adjacent superpixel block is higher, and then the pixel point is distributed to the superpixel with real defectsThe greater the probability within a block.
Finally, the defect is segmented by adopting the SLIC superpixel segmentation method in the embodiment, so that the superpixel segmentation process is the same as the conventional SLIC segmentation process, but the method is improved based on the original algorithm, and the self-adaption of the weight value is realized. The gradient values of all pixel points are calculated in a possible defect area window, a gradient minimum value point is selected as a self-adaptive segmentation seed point of the current possible defect area, and the gradient minimum value point is the pixel point with the smoothest gray level in the whole possible defect area window, so that the super-pixel segmentation speed can be increased when the super-pixel segmentation is carried out by taking the gradient minimum value point as the seed point.
For each search window pixel point in the possible defect area, calculating the distance between the pixel point and the segmentation seed point around the pixel point, and expressing the distance as follows by a formula:
wherein, the first and the second end of the pipe are connected with each other,
the distance from the pixel point b to the segmentation seed point is represented,
which represents the gray value of the pixel point b,
representing the gray value of the segmentation seed point z,
a gray-scale distance weight value is represented,
the maximum gray scale distance in the possible defect area, i.e. the difference between the maximum value and the minimum value of the gray scale value in the area, needs to be obtained by the implementer according to the actual situation,
a location distance weight value is represented,
and
respectively representing the pixel coordinates of the pixel point b and the segmentation seed point z,
the maximum spatial position distance is represented, and the value in this embodiment is the size S of the super-pixel block, which can be set by an implementer according to the actual situation.
The original distance can be accurately segmented by the weight value, so that each pixel point can be segmented to an optimal seed point region as far as possible.
And finally allocating the pixel points to the super pixel block where the seed point with the minimum distance d to the pixel points is located, and finally allocating the pixel points in each possible defect area to the proper seed point center. Setting the pixel value of the pixel point in the area divided by the super pixel to be 1, setting the pixel values of the pixel points in other areas to be 0, further obtaining a mask image, multiplying the mask image and the original preferred road image to obtain a defect analysis image, dividing the defect analysis image to obtain a defect area, identifying the defect type according to the defect area, and further judging whether the road surface quality is qualified according to the defect type. The implementer can select a suitable method for determination according to actual situations, for example, select a manual method for determination of recognition, or use a neural network for determination of recognition, which is not described in more detail in the prior art.
It should be noted that, in the present invention, the texture feature value is obtained through the defect feature matrix corresponding to the possible defect region, and then the size and distance weight parameter of the super-pixel segmentation are obtained, so that the parameters can be set adaptively according to the texture feature of the defect when the super-pixel segments the defect region, thereby improving the segmentation efficiency and also ensuring the segmentation accuracy.
The above-mentioned embodiments are only used for illustrating the technical solutions of the present application, and not for limiting the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; the modifications or substitutions do not make the essence of the corresponding technical solutions deviate from the technical solutions of the embodiments of the present application, and are included in the protection scope of the present application.
Claims (8)
1. A road quality identification detection method is characterized by comprising the following steps:
acquiring a road surface gray image, segmenting the road surface gray image to acquire a lane line region, and assigning a pixel value of a pixel point in the lane line region by using a gray value with the highest frequency in the road surface gray image to obtain an optimal road image;
dividing the pixel values of the pixel points in the preferred road image into gray levels, processing the preferred road image by using a window with a set size, and calculating the gray uniformity degree corresponding to the window according to the gray levels of the pixel points in the window; determining a possible defect area according to the gray level uniformity degree corresponding to the window;
skeleton refinement is carried out on each possible defect area, hough line detection is carried out on each possible defect area after skeleton refinement, and the defect possibility of pixel points is obtained according to Hough line detection voting values of the pixel points in the possible defect areas; obtaining a defect probability value corresponding to the gray level according to the mean value of the defect possibilities of the pixel points belonging to the same gray level;
constructing a defect feature matrix according to defect probability values corresponding to the gray levels of the pixel points, obtaining energy feature values according to the defect probability values in the defect feature matrix, and determining the size of the super-pixel block according to the energy feature values; dividing the possible defect area by using a superpixel division method according to the size of the superpixel block to obtain a defect area; and determining the defect type according to the defect area, and further identifying the road quality to obtain a quality identification result.
2. The road quality identification and detection method according to claim 1, wherein the step of segmenting the road surface grayscale image to obtain the lane line region specifically comprises:
and (3) counting gray values of pixel points in the gray image on the road surface to construct a gray histogram, determining a gray range according to the gray histogram, and selecting the pixel points with the gray values in the gray range as initial seed points to perform region growth to obtain a lane line region.
3. The road quality identification and detection method according to claim 1, wherein the dividing of the gray scale for the pixel values of the pixel points in the preferred road image specifically comprises:
the gray value of a pixel point in the optimized road image is counted to construct a gray histogram, the gray histogram is segmented by using a multi-threshold segmentation method, the gray values segmented in the same interval are recorded as the same gray level, the gray levels are sequenced according to the sequence of the gray values from small to large, and the sequence numbers are recorded as the corresponding gray levels according to the sequence.
4. The road quality identification and detection method according to claim 1, wherein the method for obtaining the degree of uniformity of the gray level corresponding to the window specifically comprises:
wherein HD represents the gray level uniformity degree corresponding to the window,
represents the gray scale of the a-th pixel point,
representing the total number of pixels in the window, exp () represents an exponential function with a natural constant e as the base.
5. The road quality identification and detection method according to claim 1, wherein the determining of the possible defect area according to the degree of uniformity of the gray scale corresponding to the window specifically comprises:
recording the set size of the window as an initial size, mining the initial size of the window according to a set step length when the gray level uniformity degree corresponding to the window of the initial size is smaller than or equal to a degree threshold, calculating the gray level uniformity degree corresponding to the expanded window, continuing to expand the size of the window when the gray level uniformity degree corresponding to the expanded window is still smaller than or equal to the degree threshold, and recording the area in the window as a possible defect area when the gray level uniformity degree corresponding to the expanded window is larger than the degree threshold.
6. The road quality identification and detection method according to claim 1, wherein the constructing of the defect feature matrix according to the defect probability values corresponding to the gray levels of the pixels specifically comprises:
and constructing a gray level co-occurrence matrix corresponding to a possible defect area according to the gray level of the pixel point, and constructing a defect feature matrix according to the defect probability value corresponding to the gray level of the pixel point and elements in the gray level co-occurrence matrix.
7. The method as claimed in claim 1, wherein the determining the size of the super-pixel block according to the energy feature value is specifically:
wherein S is the size of the super pixel block, ma represents the energy characteristic value corresponding to the defect characteristic matrix,
represents a constant coefficient, e is a natural constant,
as a function of the rounding-up.
8. The method for identifying and detecting road quality according to claim 1, wherein the segmenting the possible defect area according to the size of the super pixel block by using the super pixel segmentation method specifically comprises:
determining a search range during superpixel segmentation according to the size of the superpixel block; determining segmentation seed points according to gradient values of pixel points in a possible defect region, calculating an inverse difference moment of a defect characteristic matrix, and determining a weighted value corresponding to a distance according to the inverse difference moment; calculating the distance between the pixel point and the segmentation seed point in the corresponding search range according to the weight value, the gray value of the pixel point in the possible defect region and the gray value of the segmentation seed point, and the pixel coordinates of the pixel point in the possible defect region and the segmentation seed point; and performing super-pixel segmentation according to the distance.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211417318.0A CN115457041B (en) | 2022-11-14 | 2022-11-14 | Road quality identification and detection method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211417318.0A CN115457041B (en) | 2022-11-14 | 2022-11-14 | Road quality identification and detection method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115457041A true CN115457041A (en) | 2022-12-09 |
| CN115457041B CN115457041B (en) | 2023-02-14 |
Family
ID=84295394
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211417318.0A Active CN115457041B (en) | 2022-11-14 | 2022-11-14 | Road quality identification and detection method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115457041B (en) |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115862006A (en) * | 2023-03-01 | 2023-03-28 | 山东长有面粉有限公司 | Method for detecting bran stars in flour milling process |
| CN115909256A (en) * | 2023-01-06 | 2023-04-04 | 济宁市百卉农林发展有限公司 | Road disease detection method based on road visual image |
| CN116092015A (en) * | 2023-04-06 | 2023-05-09 | 安徽乾劲企业管理有限公司 | Road construction state monitoring method |
| CN116228765A (en) * | 2023-05-08 | 2023-06-06 | 济宁市健达医疗器械科技有限公司 | Mask ear strap welding quality detection method |
| CN116452580A (en) * | 2023-06-13 | 2023-07-18 | 山东古天电子科技有限公司 | Notebook appearance quality detection method |
| CN116485801A (en) * | 2023-06-26 | 2023-07-25 | 山东兰通机电有限公司 | Rubber tube quality online detection method and system based on computer vision |
| CN116503404A (en) * | 2023-06-27 | 2023-07-28 | 梁山县创新工艺品股份有限公司 | Plastic toy quality detection method and device, electronic equipment and storage medium |
| CN116664584A (en) * | 2023-08-02 | 2023-08-29 | 东莞市旺佳五金制品有限公司 | Intelligent feedback regulating system for production of thin-wall zinc alloy die casting die |
| CN116681703A (en) * | 2023-08-03 | 2023-09-01 | 杭州鸿世电器股份有限公司 | Intelligent switch quality rapid detection method |
| CN116703251A (en) * | 2023-08-08 | 2023-09-05 | 德润杰(山东)纺织科技有限公司 | Rubber ring production quality detection method based on artificial intelligence |
| CN116758059A (en) * | 2023-08-10 | 2023-09-15 | 吉林交通职业技术学院 | Visual nondestructive testing method for roadbed and pavement |
| CN116758065A (en) * | 2023-08-14 | 2023-09-15 | 山东嘉利诺新材料科技有限公司 | Rapid detection method for surface defects of fireproof plate |
| CN116993947A (en) * | 2023-09-26 | 2023-11-03 | 光谷技术有限公司 | Visual display method and system for three-dimensional scene |
| CN117095362A (en) * | 2023-10-20 | 2023-11-21 | 中国科学院苏州生物医学工程技术研究所 | Liquid drop monitoring method, system and storage medium of cell sorter |
| CN117173661A (en) * | 2023-11-02 | 2023-12-05 | 中铁五局集团成都工程有限责任公司 | Asphalt road quality detection method based on computer vision |
| CN117541605A (en) * | 2024-01-09 | 2024-02-09 | 山东华中重钢有限公司 | Rapid segmentation method for rusted image area of steel structure |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018153304A1 (en) * | 2017-02-22 | 2018-08-30 | 武汉极目智能技术有限公司 | Map road mark and road quality collection apparatus and method based on adas system |
| CN109115785A (en) * | 2018-08-08 | 2019-01-01 | 长沙理工大学 | A kind of casting grinding quality determining method, device and its application method |
| CN110119687A (en) * | 2019-04-17 | 2019-08-13 | 浙江工业大学 | Detection method based on the road surface slight crack defect that image procossing and convolutional neural networks combine |
-
2022
- 2022-11-14 CN CN202211417318.0A patent/CN115457041B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018153304A1 (en) * | 2017-02-22 | 2018-08-30 | 武汉极目智能技术有限公司 | Map road mark and road quality collection apparatus and method based on adas system |
| CN109115785A (en) * | 2018-08-08 | 2019-01-01 | 长沙理工大学 | A kind of casting grinding quality determining method, device and its application method |
| CN110119687A (en) * | 2019-04-17 | 2019-08-13 | 浙江工业大学 | Detection method based on the road surface slight crack defect that image procossing and convolutional neural networks combine |
Cited By (29)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115909256A (en) * | 2023-01-06 | 2023-04-04 | 济宁市百卉农林发展有限公司 | Road disease detection method based on road visual image |
| CN115909256B (en) * | 2023-01-06 | 2023-05-19 | 济宁市百卉农林发展有限公司 | Road disease detection method based on road visual image |
| CN115862006A (en) * | 2023-03-01 | 2023-03-28 | 山东长有面粉有限公司 | Method for detecting bran stars in flour milling process |
| CN116092015A (en) * | 2023-04-06 | 2023-05-09 | 安徽乾劲企业管理有限公司 | Road construction state monitoring method |
| CN116228765A (en) * | 2023-05-08 | 2023-06-06 | 济宁市健达医疗器械科技有限公司 | Mask ear strap welding quality detection method |
| CN116452580A (en) * | 2023-06-13 | 2023-07-18 | 山东古天电子科技有限公司 | Notebook appearance quality detection method |
| CN116452580B (en) * | 2023-06-13 | 2023-09-01 | 山东古天电子科技有限公司 | Notebook appearance quality detection method |
| CN116485801A (en) * | 2023-06-26 | 2023-07-25 | 山东兰通机电有限公司 | Rubber tube quality online detection method and system based on computer vision |
| CN116485801B (en) * | 2023-06-26 | 2023-09-12 | 山东兰通机电有限公司 | Rubber tube quality online detection method and system based on computer vision |
| CN116503404B (en) * | 2023-06-27 | 2023-09-01 | 梁山县创新工艺品股份有限公司 | Plastic toy quality detection method and device, electronic equipment and storage medium |
| CN116503404A (en) * | 2023-06-27 | 2023-07-28 | 梁山县创新工艺品股份有限公司 | Plastic toy quality detection method and device, electronic equipment and storage medium |
| CN116664584A (en) * | 2023-08-02 | 2023-08-29 | 东莞市旺佳五金制品有限公司 | Intelligent feedback regulating system for production of thin-wall zinc alloy die casting die |
| CN116664584B (en) * | 2023-08-02 | 2023-11-28 | 东莞市旺佳五金制品有限公司 | Intelligent feedback regulating system for production of thin-wall zinc alloy die casting die |
| CN116681703A (en) * | 2023-08-03 | 2023-09-01 | 杭州鸿世电器股份有限公司 | Intelligent switch quality rapid detection method |
| CN116681703B (en) * | 2023-08-03 | 2023-10-10 | 杭州鸿世电器股份有限公司 | Intelligent switch quality rapid detection method |
| CN116703251A (en) * | 2023-08-08 | 2023-09-05 | 德润杰(山东)纺织科技有限公司 | Rubber ring production quality detection method based on artificial intelligence |
| CN116703251B (en) * | 2023-08-08 | 2023-11-17 | 德润杰(山东)纺织科技有限公司 | Rubber ring production quality detection method based on artificial intelligence |
| CN116758059A (en) * | 2023-08-10 | 2023-09-15 | 吉林交通职业技术学院 | Visual nondestructive testing method for roadbed and pavement |
| CN116758059B (en) * | 2023-08-10 | 2023-10-20 | 吉林交通职业技术学院 | Visual nondestructive testing method for roadbed and pavement |
| CN116758065B (en) * | 2023-08-14 | 2023-10-20 | 山东嘉利诺新材料科技有限公司 | Rapid detection method for surface defects of fireproof plate |
| CN116758065A (en) * | 2023-08-14 | 2023-09-15 | 山东嘉利诺新材料科技有限公司 | Rapid detection method for surface defects of fireproof plate |
| CN116993947A (en) * | 2023-09-26 | 2023-11-03 | 光谷技术有限公司 | Visual display method and system for three-dimensional scene |
| CN116993947B (en) * | 2023-09-26 | 2023-12-12 | 光谷技术有限公司 | Visual display method and system for three-dimensional scene |
| CN117095362A (en) * | 2023-10-20 | 2023-11-21 | 中国科学院苏州生物医学工程技术研究所 | Liquid drop monitoring method, system and storage medium of cell sorter |
| CN117095362B (en) * | 2023-10-20 | 2024-02-02 | 中国科学院苏州生物医学工程技术研究所 | Liquid drop monitoring method, system and storage medium of cell sorter |
| CN117173661A (en) * | 2023-11-02 | 2023-12-05 | 中铁五局集团成都工程有限责任公司 | Asphalt road quality detection method based on computer vision |
| CN117173661B (en) * | 2023-11-02 | 2024-01-26 | 中铁五局集团成都工程有限责任公司 | Asphalt road quality detection method based on computer vision |
| CN117541605A (en) * | 2024-01-09 | 2024-02-09 | 山东华中重钢有限公司 | Rapid segmentation method for rusted image area of steel structure |
| CN117541605B (en) * | 2024-01-09 | 2024-03-29 | 山东华中重钢有限公司 | Rapid segmentation method for rusted image area of steel structure |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115457041B (en) | 2023-02-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115457041B (en) | Road quality identification and detection method | |
| CN115829883B (en) | Surface image denoising method for special-shaped metal structural member | |
| CN115311292B (en) | Strip steel surface defect detection method and system based on image processing | |
| CN114862862B (en) | Pump body cold shut defect identification method and system based on image processing | |
| CN114140462B (en) | Bearing wear degree assessment method based on image processing | |
| CN116310360B (en) | Reactor surface defect detection method | |
| CN115082418B (en) | Precise identification method for automobile parts | |
| CN109191459B (en) | Automatic identification and rating method for continuous casting billet macrostructure center segregation defect | |
| CN117173184B (en) | Road construction quality detection method and system based on artificial intelligence | |
| CN116758061B (en) | Casting surface defect detection method based on computer vision | |
| CN114219805B (en) | Intelligent detection method for glass defects | |
| CN108711158B (en) | Pointer instrument image identification method based on contour fitting and radial segmentation | |
| CN116309537B (en) | Defect detection method for oil stain on surface of tab die | |
| CN115049664B (en) | Vision-based ship engine fitting defect detection method | |
| CN114782432B (en) | Edge detection method of improved canny operator based on textural features | |
| CN115018838B (en) | Method for identifying pitting defects on surface of oxidized steel pipe material | |
| CN115311277B (en) | Pit defect identification method for stainless steel product | |
| CN114749342B (en) | Lithium battery pole piece coating defect identification method, device and medium | |
| CN111738256B (en) | Composite CT image segmentation method based on improved watershed algorithm | |
| CN115115638B (en) | Oil leakage detection and judgment method for hydraulic system | |
| CN115063430B (en) | Electric pipeline crack detection method based on image processing | |
| CN115294159B (en) | Method for dividing corroded area of metal fastener | |
| CN116385450B (en) | PS sheet wear resistance detection method based on image processing | |
| CN114782329A (en) | Bearing defect damage degree evaluation method and system based on image processing | |
| CN116883408B (en) | Integrating instrument shell defect detection method based on artificial intelligence |
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 |