CN115601630B - Stain recognition method for automatic wallboard mold cleaning machine - Google Patents
Stain recognition method for automatic wallboard mold cleaning machine Download PDFInfo
- Publication number
- CN115601630B CN115601630B CN202211609040.7A CN202211609040A CN115601630B CN 115601630 B CN115601630 B CN 115601630B CN 202211609040 A CN202211609040 A CN 202211609040A CN 115601630 B CN115601630 B CN 115601630B
- Authority
- CN
- China
- Prior art keywords
- pixel
- pixel points
- points
- abnormal
- super
- 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.)
- Active
Links
- 238000004140 cleaning Methods 0.000 title claims abstract description 37
- 238000000034 method Methods 0.000 title claims abstract description 30
- 230000002159 abnormal effect Effects 0.000 claims abstract description 125
- 238000001914 filtration Methods 0.000 claims abstract description 21
- 239000007921 spray Substances 0.000 claims abstract description 15
- 150000001875 compounds Chemical class 0.000 claims description 15
- 238000005406 washing Methods 0.000 claims description 14
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 13
- 238000004364 calculation method Methods 0.000 claims description 12
- 241001632427 Radiola Species 0.000 claims description 7
- 239000002002 slurry Substances 0.000 abstract description 96
- 239000002699 waste material Substances 0.000 abstract description 10
- 239000006185 dispersion Substances 0.000 abstract description 5
- 241001632422 Radiola linoides Species 0.000 abstract 1
- 238000012544 monitoring process Methods 0.000 description 19
- 230000011218 segmentation Effects 0.000 description 12
- 230000000694 effects Effects 0.000 description 3
- 238000007781 pre-processing Methods 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000012216 screening Methods 0.000 description 3
- 239000006082 mold release agent Substances 0.000 description 2
- 241000196324 Embryophyta Species 0.000 description 1
- 239000000853 adhesive Substances 0.000 description 1
- 230000001070 adhesive effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 239000002184 metal Substances 0.000 description 1
- 239000002861 polymer material Substances 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/36—Applying a local operator, i.e. means to operate on image points situated in the vicinity of a given point; Non-linear local filtering operations, e.g. median filtering
-
- 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
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/761—Proximity, similarity or dissimilarity measures
-
- 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
- G06V10/762—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using clustering, e.g. of similar faces in social networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Nonlinear Science (AREA)
- Image Processing (AREA)
Abstract
The invention relates to the field of image recognition, in particular to a stain recognition method for an automatic wallboard mold cleaning machine, which comprises the following steps: obtaining the density of abnormal pixel points according to the minimum achievable range and the same quantity, and obtaining all seed pixel points and corresponding superpixel blocks according to the density; calculating the space structure parameters and the gray structure parameters of the superpixel blocks, and obtaining the merging necessity of the two superpixel blocks according to the clustering labels; merging all the super-pixel blocks to obtain all the credible areas; obtaining the size of a median filtering kernel according to all direction noise points of all noise pixel points; and filtering and denoising the wallboard mould image according to the median filter kernel, and obtaining spray points according to the adhesion of the stain pixel points and cleaning. The invention accurately removes the noise pixel points in the slurry dispersion area and retains the pixel points belonging to the slurry, retains the detail information and simultaneously improves the accuracy of subsequent spot area identification, thereby avoiding resource waste caused by repeated cleaning and overall cleaning.
Description
Technical Field
The invention relates to the field of image recognition, in particular to a stain recognition method for an automatic wallboard mold cleaning machine.
Background
The novel high polymer material light wallboard needs to be poured by using a metal mold, although a mold release agent is coated on the mold before pouring, slurry is inevitably left to form stains after pouring, and the quality of the next batch of poured wallboards can be influenced if the mold release agent is not completely cleaned. Because the efficiency of pure manual cleaning is low, equipment is usually adopted for automatic cleaning, the automatic cleaning needs to be assisted by a monitoring system right above a factory building to finish cleaning operation, an image sensor carried by the monitoring system can automatically identify slurry remained on a wallboard die as a stain area, and then the cleaning equipment is controlled to flush and clean the stain area. The purpose of identifying the stain area is to accurately position the stain position and control the water gun nozzle to wash the stain area, so that resource waste caused by repeated cleaning and overall cleaning is avoided.
The generation of white noise due to long running time of the monitoring system is a common problem, so most monitoring systems include the function of image preprocessing. The monitoring image is preprocessed through a conventional noise reduction and denoising algorithm, so that the recognition degree of the monitoring image in human vision is only improved, and the method is only suitable for the auxiliary manual cleaning operation of a monitoring system; when the automatic cleaning is carried out based on the monitoring system, when the monitoring image is preprocessed through a conventional filtering and denoising algorithm, part of detail information in the monitoring image is removed together, and the accuracy of subsequent spot area identification is greatly influenced.
How to filter the monitoring image, the detail information in the monitoring image is kept while the noise is removed, the accuracy of subsequent stain area identification is improved, the position of the stain is accurately positioned, and the resource waste caused by repeated cleaning and overall cleaning is avoided.
Disclosure of Invention
In order to solve the above problems, the present invention provides a stain recognition method for an automatic wallboard mold cleaning machine, the method comprising:
acquiring all abnormal pixel points of the wallboard mold image;
for any abnormal pixel point, obtaining the minimum reachable range and the number of the same types of the abnormal pixel points according to the local reachable range radius of the abnormal pixel point, obtaining the density of the abnormal pixel points according to the minimum reachable range and the number of the same types of the abnormal pixel points, and taking the abnormal pixel points with the density larger than the density threshold value as seed pixel points; obtaining all seed pixel points, and obtaining all superpixel blocks according to the minimum reachable range of all the seed pixel points;
for any super pixel block, calculating the space structure parameter and the gray structure parameter of the super pixel block, and taking the space structure parameter and the gray structure parameter of the super pixel block as a clustering label of the super pixel block; obtaining clustering labels of all superpixel blocks;
for any two super-pixel blocks, acquiring the merging necessity degree of the two super-pixel blocks according to the clustering labels of the two super-pixel blocks; merging two super-pixel blocks with merging necessity degrees larger than a merging threshold value, merging the super-pixel blocks corresponding to all seed pixel points according to the merging necessity degrees of any two super-pixel blocks, marking the area corresponding to each merged super-pixel block as a credible area, and acquiring all credible areas on the wallboard mold image;
marking pixel points with gradient amplitudes larger than a gradient threshold value in all credible regions as noise pixel points, acquiring directional noise points of the noise pixel points in all neighborhood directions, and acquiring the size of a median filter kernel according to all directional noise points of all noise pixel points;
filtering and denoising the wallboard mould image according to the median filtering kernel of the size to obtain all stain pixel points on the denoised wallboard mould image; and calculating the adhesion degree of all the stain pixel points, taking the stain pixel points with the adhesion degrees larger than the adhesion threshold value as spray washing points, and cleaning by using a water gun spray head.
Further, the step of obtaining the minimum reachable range and the number of the same type of abnormal pixel points according to the local reachable range radius of the abnormal pixel points comprises:
two pixel points with the gray value difference not greater than 5 are taken as the same type of pixel points; respectively recording 8 neighborhood directions of the abnormal pixel points from 1 neighborhood direction to 8 neighborhood directions, and calculating the average value of the abnormal pixel points according to the average value
Searching in the neighborhood direction to obtain
The abnormal pixel point which is closest to the abnormal pixel point in the neighborhood direction and belongs to the same type of pixel point as the abnormal pixel point is marked as the abnormal pixel point
Direction homologous points; obtaining direction homogeneous points of the abnormal pixel points in 8 neighborhood directions;
taking the maximum value in all Euclidean distances between the same point in 8 directions and the abnormal pixel point as the local reachable range radius of the abnormal pixel point, and taking a circular area which takes the pixel point as the center and the local reachable range radius of the pixel point as the radius as the minimum reachable range of the abnormal pixel point;
and counting the number of all abnormal pixel points which are within the minimum reachable range of the abnormal pixel point and belong to the same type of pixel points as the abnormal pixel point, and recording the number as the same type of abnormal pixel points.
Further, the step of obtaining the density of the abnormal pixel points according to the minimum achievable range and the number of the same type of the abnormal pixel points comprises:
first, the
The calculation formula of the density of each abnormal pixel point is as follows:
in the formula (I), the compound is shown in the specification,
is shown as
The density of the individual abnormal pixel points is,
is shown as
The number of the same type of the abnormal pixel points,
denotes the first
The area of the minimum reachable range of the abnormal pixel,
is shown as
The local reachable range radius of each abnormal pixel point.
Further, the step of calculating spatial structure parameters of the superpixel block comprises:
first, the
Superpixel block corresponding to seed pixel point
Spatial structure parameter of
The calculation formula of (c) is:
in the formula (I), the compound is shown in the specification,
is shown as
Superpixel block corresponding to seed pixel point
The spatial structure parameters of (a) are,
represented in a superpixel block
In, the first
A seedOf pixels
Sum in the neighborhood direction
The seed pixel point belongs to the first of the same kind
The number of the abnormal pixel points is one,
，
representing abnormal pixel points
Is determined by the coordinate of (a) in the space,
，
is as follows
The coordinates of the pixel points of each seed,
is shown as
Of individual seed pixel points
Sum in the neighborhood direction
The number of abnormal pixels of which the seed pixels belong to the same type of pixels.
Further, the step of calculating the gray structure parameter of the super pixel block comprises:
first, the
Superpixel block corresponding to seed pixel point
Gray scale structure parameter of
The calculation formula of (2) is as follows:
in the formula (I), the compound is shown in the specification,
is shown as
Superpixel block corresponding to seed pixel point
The gray-scale structure parameter of (a),
representing superpixel blocks
The number of all the gray-scale values within,
indicating gray value equal to gray value
The number of all the pixel points of (a),
denotes the first
The number of gray-scale values is,
is shown as
The local reachable range radius of each seed pixel.
Further, the step of obtaining the merging necessity of the two super-pixel blocks according to the clustering labels of the two super-pixel blocks comprises:
for the first
Superpixel block corresponding to seed pixel point
And a first
Superpixel block corresponding to seed pixel point
Computing superpixel blocks
And superpixel block
The calculation formula of the merging necessity degree is as follows:
in the formula (I), the compound is shown in the specification,
representing superpixels
And superpixel blocks
The degree of necessity of the merging of (1),
denotes the first
Superpixel block corresponding to seed pixel point
The spatial structure parameters of (a) are,
denotes the first
Superpixel block corresponding to seed pixel point
The spatial structure parameters of (a) are,
denotes the first
Superpixel block corresponding to seed pixel point
The gray-scale structure parameter of (a),
is shown as
Superpixel block corresponding to seed pixel point
The gray-scale structure parameter of (a),
denotes the first
Seed pixel point and the second
The Euclidean distance of the pixel points of each seed,
which represents a function of the hyperbolic tangent,
representing the L2 norm.
Further, the step of obtaining the size of the median filtering kernel according to all directional noise points of all the noise pixel points includes:
the size of the median filter kernel is calculated as:
in the formula (I), the compound is shown in the specification,
the size of the median filtering kernel is indicated,
representing the number of all noisy pixels in all trusted regions,
denotes the first
Of a noise pixel point and of the noise pixel point
Directional noise point
The euclidean distance of (c).
Further, the step of calculating the attachment degrees of all the dirty pixel points includes:
acquiring a circular area which takes the spot pixel points as the center and takes the water spray radius as the radius, counting to acquire the number of all spot pixel points in the circular area, and taking the ratio of the number of all spot pixel points in the circular area to the area of the circular area as the attachment degree of the spot pixel points.
The embodiment of the invention at least has the following beneficial effects:
1. due to noise interference, the conventional method such as a superpixel segmentation algorithm cannot completely segment a slurry dense region from a wallboard mold image, seed pixel points are screened from all abnormal pixel points according to the density of the abnormal pixel points, all the seed pixel points are ensured to belong to the pixel points belonging to the slurry in the slurry dense region in the wallboard mold image, a superpixel block obtained based on the seed pixel points belongs to the slurry dense region, the slurry dense region can be completely segmented from the wallboard mold image, and the distribution characteristic of the noise pixel points in the slurry dense region can be accurately obtained.
2. The invention combines the characteristics that noise pixel points have the same distribution characteristics in a slurry dense area and the slurry dispersed area and the noise pixel points are easy to distinguish from the noise pixel points in the slurry dense area, obtains the distribution characteristics of the noise pixel points in the slurry dispersed area according to the distribution characteristics of the noise pixel points in the slurry dense area, further combines the distribution characteristics of the noise pixel points to denoise the wallboard mold image, removes the noise pixel points and retains the pixel points belonging to the slurry in the slurry dispersed area, and accurately retains the detailed information of the wallboard mold image.
3. According to the method, the median filter kernel with a proper size is obtained according to the distribution characteristics of the noise pixel points, and then the wallboard mold image is denoised through the median filter kernel, compared with the situation that the detail information of the wallboard mold image is lost due to the fact that the wallboard mold image is preprocessed through a conventional denoising method, the method removes the noise pixel points, reserves the pixel points belonging to slurry in a slurry dispersion area, accurately reserves the detail information of the wallboard mold image, improves the accuracy of subsequent stain area identification, accurately positions stain positions, obtains accurate spray washing points, and avoids resource waste caused by repeated washing and overall washing.
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 flowchart illustrating the steps of a stain recognition method for an automatic wallboard mold cleaning machine according to one embodiment of the present invention;
fig. 2 shows 8 neighborhood directions of a pixel point according to an embodiment 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 of the method for identifying stains for an automatic cleaning machine of wallboard mold according to the present invention with reference to the accompanying drawings and preferred embodiments, the detailed implementation, structure, features and effects thereof are described as follows. 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 stain recognition method for the automatic wallboard mold cleaning machine provided by the invention in detail with reference to the accompanying drawings.
Referring to fig. 1, a flowchart illustrating steps of a stain recognition method for an automatic wallboard mold cleaning machine according to an embodiment of the present invention is shown, wherein the method includes the following steps:
and S001, acquiring the wallboard mold image by using a plant monitoring system.
Install video monitor system directly over wallboard processing area, adjust indoor ambient light, diffuse reflection lamp source is chooseed for use to light, avoids the light to penetrate directly the reflection of light interference that brings.
The monitoring system is networked with the automatic cleaning machine, when the wallboard mold is cleaned, the wallboard mold image of the static frame is intercepted from the monitoring system, the preprocessing module carries out graying processing on the image, the running speed of a subsequent algorithm is accelerated, and the interference of redundant color information is eliminated.
Acquiring a standard image of the wallboard mold, calculating a peak signal-to-noise ratio of the wallboard mold image and the standard image according to the gray value of the pixel point, and judging whether the wallboard mold image needs to be denoised according to the peak signal-to-noise ratio of the wallboard mold image and the standard image. And if the peak signal-to-noise ratio of the monitoring image and the standard image is greater than a first threshold value, performing noise reduction processing on the monitoring image.
S002, obtaining all abnormal pixel points of the wallboard mold image, obtaining the minimum reachable range and the number of the same type of the abnormal pixel points according to the local reachable range radius of the abnormal pixel points, obtaining the density of the abnormal pixel points according to the minimum reachable range and the number of the same type of the abnormal pixel points, and obtaining all seed pixel points according to the density screening of the abnormal pixel points.
It should be noted that white noise is a common problem caused by long running time of the monitoring system, and therefore most monitoring systems include a function of image preprocessing. Slurry on the surface of the wallboard mould is remained, when the stains on the surface of the wallboard are identified, the stains with the largest area are covered as far as possible by the position sprayed by the water gun, so that the stains can be efficiently removed and cleaning resources are saved, and in practice, the slurry on the wallboard mould is in an existing dense state and a dispersed state, namely, a slurry dense area and a slurry dispersed area exist in the image of the wallboard mould.
The pixel points belonging to the slurry and the noise pixel points in the slurry dispersing area are distributed in a relatively dispersed manner, so that the pixel points belonging to the slurry and the noise pixel points in the slurry dispersing area are difficult to distinguish, the pixel points belonging to the slurry in the slurry dispersing area are filtered by a conventional noise reduction method, the detail information of the wallboard mould image is lost, the subsequent identification accuracy of the stain area is low, the stain position cannot be accurately positioned, and the resource waste caused by repeated cleaning and global cleaning is caused.
Considering that the distribution of the pixel points belonging to the slurry in the slurry dense region is dense, that is, the pixel points belonging to the slurry in the slurry dense region are connected into one piece, and the distribution characteristics of the noise pixel points are different from those of the noise pixel points, the distribution characteristics of the noise pixel points can be more accurately obtained in the slurry dense region. The white noise has the characteristic of random uniform discrete distribution on the image, so that noise pixel points have the same distribution characteristics in a slurry dense area and a slurry dispersed area, the distribution characteristics of the noise pixel points in the slurry dispersed area can be obtained according to the distribution characteristics of the noise pixel points in the slurry dense area, the noise of the wallboard mould image is removed by combining the distribution characteristics of the noise pixel points, the pixel points belonging to the slurry in the slurry dispersed area are reserved by removing the noise pixel points, the detail information of the wallboard mould image is accurately reserved, the accuracy of subsequent spot area identification is improved, the spot position is accurately positioned, and the resource waste caused by repeated cleaning and global cleaning is avoided.
Filtering the wallboard mould image through median filtering, and reserving pixel points belonging to slurry in a slurry dispersion area in order to accurately remove noise pixel points; however, due to noise interference, the slurry dense region cannot be completely segmented from the wallboard mold image directly through threshold segmentation, so that the method considers that the slurry dense region in the wallboard mold image is obtained through a superpixel segmentation algorithm, and noise pixel points interfere with the acquisition of initial seed points and the setting of clustering step length in the superpixel segmentation algorithm.
In order to avoid setting the initial seed points in the superpixel segmentation algorithm, the noise pixel points or the pixel points belonging to the slurry in the slurry dispersion area are set as the initial seed points, and then abnormal superpixel blocks appear, the invention screens the seed pixel points from all the abnormal pixel points according to the density of the abnormal pixel points, ensures that all the seed pixel points belong to the pixel points belonging to the slurry in the slurry dense area in the wallboard mold image, and further ensures that the superpixel blocks obtained based on the seed pixel points belong to the slurry dense area.
In this embodiment, the specific steps of obtaining all the abnormal pixel points, calculating the density of the abnormal pixel points, and screening and obtaining the seed pixel points according to the density of the abnormal pixel points are as follows:
1. and acquiring abnormal pixel points.
And calculating the difference value of the gray values of the pixel points at the same positions in the wallboard mould image and the standard image, and recording the pixel points with the difference value not being 0 in the wallboard mould image as abnormal pixel points.
2. And obtaining the minimum reachable range and the same type number of the abnormal pixel points according to the local reachable range radius of the abnormal pixel points.
Considering that small brightness difference exists between similar pixel points on the wallboard mold image, two pixel points with the gray value difference not greater than the gray threshold value are set as the similar pixel points; for any abnormal pixel point, there are 8 neighborhood directions as shown in fig. 2, which are recorded as 1 neighborhood direction to 8 neighborhood directions, respectively.
In this embodiment, the grayscale threshold is 5, and in other embodiments, the implementer can set the grayscale threshold as needed.
For abnormal pixel point according to
Searching in the neighborhood direction to obtain
The abnormal pixel point which is closest to the abnormal pixel point in the neighborhood direction and belongs to the same type of pixel point as the abnormal pixel point is marked as the abnormal pixel point
Direction homologous points; obtaining direction homologous points of the abnormal pixel points in 8 neighborhood directions;
taking the maximum value in all Euclidean distances between the same type points in 8 directions and the abnormal pixel point as the local reachable range radius of the abnormal pixel point, taking a circular area which takes the pixel point as the center and the local reachable range radius of the pixel point as the radius as the minimum reachable range of the abnormal pixel point, counting and obtaining the number of all abnormal pixel points which are within the minimum reachable range of the abnormal pixel point and belong to the same type pixel point with the abnormal pixel point, and recording the number as the same type number of the abnormal pixel point.
3. And obtaining the density of the abnormal pixel points according to the minimum achievable range and the same type number of the abnormal pixel points, and screening according to the density of the abnormal pixel points to obtain all seed pixel points.
Obtaining the density of the abnormal pixel points according to the minimum reachable range of the abnormal pixel points and the number of the same types of the abnormal pixel points, wherein the specific calculation formula is as follows:
in the formula (I), the compound is shown in the specification,
is shown as
The density of the individual abnormal pixel points is,
is shown as
The number of the same type of the abnormal pixel points,
is shown as
The area of the minimum reachable range of the abnormal pixel,
is shown as
The local reachable range radius of each abnormal pixel point.
If the density of the abnormal pixel points is larger than the density threshold value, the minimum reachable range of the abnormal pixel points belongs to the core area of the slurry dense area in the wallboard mold image, namely the core area of the area with the slurry pixel points densely distributed, otherwise, the minimum reachable range of the abnormal pixel points belongs to the edge area or the slurry dispersed area in the wallboard mold image, namely the area with the slurry pixel points dispersedly distributed, and therefore, the abnormal pixel points with the density larger than the density threshold value are used as seed pixel points to obtain all the seed pixel points.
In this embodiment, the density threshold is 0.75, and in other embodiments, the practitioner can set the density threshold as desired.
Compared with the conventional superpixel segmentation algorithm, the superpixel segmentation algorithm is suitable for the wallboard mold image with the existing noise, the seed pixel points are screened from all abnormal pixel points according to the density of the abnormal pixel points, all the seed pixel points belong to the pixel points belonging to the slurry in the slurry dense area in the wallboard mold image, the superpixel blocks are obtained based on the seed pixel points and belong to the slurry dense area, the slurry dense area can be completely segmented from the wallboard mold image, and the distribution characteristics of the noise pixel points in the slurry dense area can be accurately obtained.
Step S003, all the superpixel blocks corresponding to all the seed pixel points are obtained, the space structure parameters and the gray structure parameters of the superpixel blocks are calculated, and the clustering labels of all the superpixel blocks are obtained; and obtaining the merging necessity degree according to the clustering label of the super pixel block, merging the super pixel block according to the merging necessity degree, and obtaining all credible areas.
1. And obtaining all superpixel blocks corresponding to all seed pixel points.
It should be noted that, in the conventional superpixel segmentation, clustering is performed by using a single pixel point as a basic unit, but noise existing on a wallboard mold image causes the result of superpixel segmentation to have the characteristics of large coverage and poor connectivity, and a complete slurry dense area on the wallboard mold image cannot be obtained. Therefore, the invention obtains the superpixel blocks through the seed pixel points distributed in the slurry dense area, and ensures that the superpixel blocks belong to the complete slurry dense area on the wallboard mould image.
In the present embodiment, for
The seed pixel point is to be
Each seed pixel point is a center and has a size of
As the area of
Superpixel block corresponding to seed pixel point
Wherein, in the step (A),
is shown as
The local reachable range radius of each abnormal pixel point.
2. And calculating the spatial structure parameters and the gray structure parameters of the superpixel blocks to obtain the clustering labels of all the superpixel blocks.
It should be noted that slurry pixel points and noise pixel points only exist in the region where the slurry pixel points are densely sliced, and slurry, background and noise pixel points exist in the region where the slurry pixel points are distributed and dispersed, so that the gray value structures of the two regions are different.
In this embodiment, when the seed pixel points are obtained in step S002, it is ensured that the minimum reachable range of all the seed pixel points belongs to the core region of the slurry-dense region in the wallboard mold image, that is, the core region of the region in which the slurry pixel points are densely distributed, and therefore, for the first time, the minimum reachable range of all the seed pixel points belongs to the core region of the slurry-dense region in the wallboard mold image
Superpixel block corresponding to seed pixel point
Can be based on superpixel blocks
Inner grey value structure to obtain superpixel blocks
The number of the clustering labels of the super-pixel block is two, and the two clustering labels are respectively a space structure parameter and a gray structure parameter.
First, the
Superpixel block corresponding to seed pixel point
Spatial structure parameter of
The calculation formula of (2) is as follows:
in the formula (I), the compound is shown in the specification,
is shown as
Superpixel block corresponding to seed pixel point
The spatial structure parameters of (a) are,
represented in superpixel blocks
In, the first
Of individual seed pixel points
Sum in the neighborhood direction
The seed pixel point belongs to the first of the same kind
The abnormal pixel points are selected from the group of abnormal pixel points,
，
representing abnormal pixel points
The coordinates of (a) are calculated,
，
is as follows
The coordinates of the individual seed pixel points,
is shown as
Of individual seed pixel points
Sum in the neighborhood direction
The number of abnormal pixels of which the seed pixel belongs to the same type of pixels,
represented in superpixel blocks
In, the first
Sum of all neighborhood directions of seed pixel point
All abnormal images of seed pixel points belonging to the same type of pixel pointsThe number of prime points.
Obtaining superpixel blocks
All gray values in the interior, in common
A gray value of
The individual gray scale value is recorded as
(ii) a Obtaining the first from the structure of all gray values
Superpixel block corresponding to seed pixel point
A gray scale structure parameter of
Superpixel block corresponding to seed pixel point
Gray scale structure parameter of
The calculation formula of (2) is as follows:
in the formula (I), the compound is shown in the specification,
is shown as
A seed pixel point pairCorresponding superpixel block
The gray-scale structure parameter of (a),
representing superpixel blocks
The number of all the gray-scale values within,
indicating gray value equal to gray value
The number of all the pixel points of (c),
denotes the first
A gray-scale value is calculated from the gray-scale value,
is shown as
The local reachable range radius of each seed pixel.
To superpixel blocks
Spatial structure parameter of
And gray scale structure parameter
As superpixels
The cluster label of (1).
3. And obtaining the merging necessity degree according to the clustering label of the superpixel block.
For the first
Superpixel block corresponding to seed pixel point
And a first
Superpixel block corresponding to seed pixel point
Computing superpixel blocks
And superpixel block
The calculation formula of the merging necessity degree is as follows:
in the formula (I), the compound is shown in the specification,
representing superpixels
And superpixel block
The degree of necessity of the merging of (1),
denotes the first
Superpixel block corresponding to seed pixel point
The spatial structure parameters of (a) are,
denotes the first
Superpixel block corresponding to seed pixel point
The spatial structure parameters of (a) are,
is shown as
Superpixel block corresponding to seed pixel point
The gray-scale structure parameter of (a),
is shown as
Superpixel block corresponding to seed pixel point
The gray-scale structure parameter of (a),
is shown as
A seed pixel point and a second seed pixel point
The Euclidean distance of the pixel points of each seed,
which represents a function of the hyperbolic tangent,
representing the L2 norm.
For superpixel blocks
And superpixel block
The more similar and closer the clustering labels of the two superpixel blocks, the more likely the two superpixel blocks belong to the same slurry dense region in the wallboard mold image, and the more necessary the two superpixel blocks need to be merged. The clustering label of the super-pixel block comprises a space structure parameter and a gray structure parameter, and the super-pixel block
And superpixel block
The more similar the spatial structure parameters are, the
The closer to 0, the superpixel block
And superpixel block
The more similar the gray scale structure parameters are, the
The closer to 0, the
Seed pixel point and the second
The closer the distance between the seed pixel points is, the Euclidean distance between the two seed pixel points
The smaller; therefore, the number of the first and second electrodes is increased,
、
and
the smaller the cluster labels of the two superpixels are, the more similar and the closer the distance is, the superpixel block
And superpixel blocks
The greater the degree of merging necessity of (c).
4. And merging the superpixel blocks according to the merging necessity degree to obtain all credible areas.
If the merging necessity degree of the two superpixel blocks is larger than the merging threshold value, the two superpixel blocks belong to the same slurry dense region in the wallboard mold image, and the two superpixel blocks are merged, otherwise, the two superpixel blocks do not belong to the same slurry dense region in the wallboard mold image, and the two superpixel blocks are not merged. It should be noted that the merge operation is to cancel a common edge between two superpixels to be merged, so that the two superpixels are merged into one superpixel block.
In this embodiment, the merge threshold is 0.8, and in other embodiments, the implementer may set the merge threshold as needed.
And combining the super pixel blocks corresponding to all the seed pixel blocks according to the combination necessity of any two super pixel blocks, recording the area corresponding to each combined super pixel block as a credible area, and obtaining all credible areas on the wallboard mould image.
Compared with the conventional superpixel segmentation algorithm, the superpixel segmentation algorithm is suitable for the wallboard mold image with the existing noise, ensures that the segmented superpixel blocks are in the core area of the slurry dense area, and obtains the credible area by combining the superpixel blocks with the same space structure parameters and gray structure parameters, so that the credible area has larger connectivity, and finally the noise distribution characteristics obtained based on the credible area have higher reliability.
And step S004, obtaining all noise pixel points of the credible area, obtaining the size of a median filter kernel according to all the noise pixel points, filtering the wallboard mould image according to the median filter kernel, calculating the adhesion of the dirty pixel points, and obtaining all spray washing points.
1. And obtaining all noise pixel points of the credible region.
And obtaining gradient amplitudes of all pixel points in the credible region by using the sobel operator, recording the pixel points with the gradient amplitudes larger than the gradient threshold value as noise pixel points of the credible region, and obtaining all noise pixel points of the credible region.
In this embodiment, the gradient threshold is 10, and in other embodiments, the practitioner may set the gradient threshold as desired.
The method combines the characteristics that noise pixel points have the same distribution characteristics in a slurry dense area and a slurry dispersed area and the characteristics that the noise pixel points are easy to distinguish from the noise pixel points in the slurry dense area, obtains the distribution characteristics of the noise pixel points in the slurry dispersed area according to the distribution characteristics of the noise pixel points in the slurry dense area, further combines the distribution characteristics of the noise pixel points to remove noise from the wallboard mold image, removes the noise pixel points, retains the pixel points belonging to the slurry in the slurry dispersed area, and accurately retains the detail information of the wallboard mold image.
2. And obtaining the size of the median filtering kernel according to all the noise pixel points.
For any noise pixel point, 8 neighborhood directions shown in fig. 2 are recorded as 1 neighborhood direction to 8 neighborhood directions respectively, and for the noise pixel point, the pixel points are arranged according to the following formula
Searching in the neighborhood direction to obtain
The noise pixel point closest to the abnormal pixel point in the neighborhood direction is recorded as the noise pixel point
Directional noise point
(ii) a And obtaining direction noise points of the noise pixel points in 8 neighborhood directions.
The size of the median filter kernel is calculated as:
in the formula (I), the compound is shown in the specification,
the size of the median filtering kernel is indicated,
representing the number of all noisy pixels in all trusted regions,
is shown as
Of a noise pixel point and of the noise pixel point
Directional noise point
The euclidean distance of (c).
The noise points are distributed on the whole image uniformly, so that the median filtering kernel is suitable for the whole image, and the smoothing effect of noise is restrained so far, namely the median filtering is self-adaptive filtering obtained according to the noise distribution on the image, and for pixel points with other distribution characteristics, such as discretely distributed slurry pixel points, the noise points can be reserved to a certain extent.
3. And filtering the wallboard mould image according to the median filtering kernel, calculating the attachment degree of the stain pixel points, and obtaining all spray-washing points.
Filtering and denoising the wallboard mould image according to the median filter kernel to obtain a denoised wallboard mould image; performing threshold segmentation on the denoised wallboard mold image to obtain all stain pixel points on the wallboard mold image; the method comprises the steps of obtaining the water spray radius of a water gun, traversing all spot pixel points, obtaining a circular area which takes the spot pixel points as the center and the water spray radius as the radius for any spot pixel point, counting the number of all spot pixel points in the circular area, and taking the ratio of the number of all spot pixel points in the circular area to the area of the circular area as the attachment degree of the spot pixel points.
The adhesion degree of the stain pixel points is larger, which shows that the stains in the areas where the stain pixel points are located are more and more concentrated, the adhesive force of the slurry is stronger, and the slurry needs to be cleaned through a water gun nozzle; the adhesion degree of the stain pixel points is small, which shows that the stains in the areas where the stain pixel points are located are less, the distribution is more dispersed, the adhesion force of the slurry is weak, and the cleaning is carried out by utilizing the attached water flow; therefore, the stain pixel points with the attachment degree larger than the attachment threshold are used as spray points and are cleaned by the water gun spray head, and the stain pixel points with the attachment degree smaller than the attachment threshold are not used as the spray points and are cleaned by the attached water flow, so that unnecessary waste of cleaning resources is reduced.
In this embodiment, the attachment threshold is 0.2, and in other embodiments, the implementer may set the merge threshold as needed.
According to the method, the median filter kernel with a proper size is obtained according to the distribution characteristics of the noise pixel points, and then the wallboard mold image is denoised through the median filter kernel, compared with the situation that the detail information of the wallboard mold image is lost due to the fact that the wallboard mold image is preprocessed through a conventional denoising method, the method removes the noise pixel points, reserves the pixel points belonging to slurry in a slurry dispersion area, accurately reserves the detail information of the wallboard mold image, improves the accuracy of subsequent stain area identification, accurately positions stain positions, obtains accurate spray washing points, and avoids resource waste caused by repeated washing and overall washing.
According to the invention, the median filter kernel with a proper size is obtained according to the distribution characteristics of the noise pixel points in the slurry dense area, and the noise pixel points have the same distribution characteristics in the slurry dense area and the slurry dispersed area, so that the wallboard mold image is denoised according to the median filter kernel, the noise pixel points are removed, the pixel points belonging to the slurry in the slurry dispersed area are reserved, the detail information of the wallboard mold image is accurately reserved, the accuracy of subsequent stain area identification is improved, the stain position is accurately positioned, and the resource waste caused by repeated cleaning and overall cleaning is avoided.
In conclusion, the method screens the seed pixel points from all the abnormal pixel points according to the density of the abnormal pixel points, ensures that the superpixel blocks obtained based on the seed pixel points belong to the slurry dense area, and can completely divide the slurry dense area from the wallboard mold image; the distribution characteristics of noise pixel points in a slurry dispersed area are obtained according to the distribution characteristics of the noise pixel points in the slurry dense area, and then the distribution characteristics of the noise pixel points are combined to denoise the wallboard mould image, remove the noise pixel points, reserve the pixel points belonging to the slurry in the slurry dispersed area, and accurately reserve the detail information of the wallboard mould image; the method includes the steps that a median filtering kernel with a proper size is obtained according to distribution characteristics of noise pixel points, then the wallboard mould image is subjected to denoising through the median filtering kernel, noise pixel points are removed, pixel points belonging to slurry in a slurry dispersing area are reserved, detail information of the wallboard mould image is accurately reserved, accuracy of subsequent stain area identification is improved, stain positions are accurately located, accurate spray-washing points are obtained, and resource waste caused by repeated washing and overall washing is avoided.
Claims (5)
1. A stain recognition method for an automatic wallboard mold cleaning machine, the method comprising:
acquiring all abnormal pixel points of the wallboard mold image;
for any abnormal pixel point, obtaining the minimum reachable range and the number of the same types of the abnormal pixel points according to the local reachable range radius of the abnormal pixel point, obtaining the density of the abnormal pixel points according to the minimum reachable range and the number of the same types of the abnormal pixel points, and taking the abnormal pixel points with the density larger than a density threshold value as seed pixel points; obtaining all seed pixel points, and obtaining all superpixel blocks according to the minimum reachable range of all the seed pixel points;
for any super pixel block, calculating the space structure parameter and the gray structure parameter of the super pixel block, and taking the space structure parameter and the gray structure parameter of the super pixel block as a clustering label of the super pixel block; obtaining clustering labels of all superpixel blocks;
for any two super-pixel blocks, acquiring the merging necessity degree of the two super-pixel blocks according to the clustering labels of the two super-pixel blocks; merging two super-pixel blocks with merging necessity degrees larger than a merging threshold value, merging the super-pixel blocks corresponding to all seed pixel points according to the merging necessity degrees of any two super-pixel blocks, marking the area corresponding to each merged super-pixel block as a credible area, and acquiring all credible areas on the wallboard mold image;
marking pixel points with gradient amplitudes larger than a gradient threshold value in all credible regions as noise pixel points, acquiring directional noise points of the noise pixel points in all neighborhood directions, and acquiring the size of a median filter kernel according to all directional noise points of all noise pixel points;
filtering and denoising the wallboard mould image according to the median filter kernel of the size to obtain all stain pixel points on the denoised wallboard mould image; calculating the attachment degrees of all the stain pixel points, taking the stain pixel points with the attachment degrees larger than an attachment threshold value as spray washing points, and washing by using a water gun spray head;
the step of calculating spatial structure parameters of a superpixel block comprises:
first, the
Super pixel block corresponding to seed pixel points>
Is based on the spatial structure parameter->
The calculation formula of (2) is as follows:
in the formula (I), the compound is shown in the specification,
indicates the fifth->
Super pixel block corresponding to each seed pixel point->
Is based on the spatial structure parameter of (4)>
Represented in the super pixel block->
Inner and second judgment>
Whether or not a seed pixel point is selected>
In the direction of the neighborhood and/or>
The fifth or fifth seed pixel belongs to the same type of pixel>
An abnormal pixel point>
，
Represents an abnormal pixel point->
Is greater than or equal to>
，
Is the first->
Coordinates of each seed pixel point>
Indicates the fifth->
On seed pixel point>
In the direction of the neighborhood and/or>
The number of abnormal pixel points of which the seed pixel points belong to the same type of pixel points;
the step of calculating the gray structure parameters of the super pixel block comprises:
first, the
Super pixel block corresponding to each seed pixel point->
Gray scale structure parameter of>
The calculation formula of (2) is as follows:
in the formula (I), the compound is shown in the specification,
indicates the fifth->
Super pixel block corresponding to each seed pixel point->
Is based on the gray scale structure parameter of (4)>
Represents a super pixel block pick>
Number of all gray values in>
Representing a gray value equal to gray value +>
Is greater than or equal to the number of all pixel points>
Indicates the fifth->
In a respective gray value +>
Indicates the fifth->
The local reachable range radius of each seed pixel point; />
The step of obtaining the merging necessity of the two superpixel blocks according to the clustering labels of the two superpixel blocks comprises the following steps:
for the first
Super pixel block corresponding to each seed pixel point->
And a fifth->
Super pixel block corresponding to each seed pixel point->
Calculating a super pixel block->
And a super pixel block>
The calculation formula of the merging necessity degree is as follows:
in the formula (I), the compound is shown in the specification,
representing a super pixel block>
And the super pixel block->
Is combined with the necessity degree of (4), (4)>
Represents a fifth or fifth party>
Super pixel block corresponding to seed pixel points>
Is based on the spatial structure parameter of (4)>
Indicates the fifth->
Super pixel block corresponding to each seed pixel point->
The spatial structure parameters of (a) are,
indicates the fifth->
Super pixel block corresponding to each seed pixel point->
Is based on the gray scale structure parameter of (4)>
Indicates the fifth->
Super pixel block corresponding to each seed pixel point->
In the gray scale structure parameter of (1), based on the gray scale structure parameter of the image signal>
Indicates the fifth->
The seed pixel point and the ^ th->
The Euclidean distance of the pixel points of each seed,
represents a hyperbolic tangent function, is selected>
Representing the L2 norm.
2. The stain recognition method for the wallboard mold automatic cleaner according to claim 1, wherein the step of obtaining the minimum reachable range and the number of the same kind of abnormal pixel points according to the local reachable range radius of the abnormal pixel points comprises:
taking two pixel points with the gray value difference not greater than the gray threshold value as the same type of pixel points; respectively recording 8 neighborhood directions of the abnormal pixel points from 1 neighborhood direction to 8 neighborhood directions, and calculating the average value of the abnormal pixel points according to the average value
Searching in the neighborhood direction to obtain a value of->
The abnormal pixel point which is closest to the abnormal pixel point in the neighborhood direction and belongs to the same type of pixel point as the abnormal pixel point is recorded as the abnormal pixel point>
Direction homogeneous point(ii) a Obtaining direction homogeneous points of the abnormal pixel points in 8 neighborhood directions;
taking the maximum value in all Euclidean distances between the same point in 8 directions and the abnormal pixel point as the local reachable range radius of the abnormal pixel point, and taking a circular area which takes the pixel point as the center and the local reachable range radius of the pixel point as the radius as the minimum reachable range of the abnormal pixel point;
and counting the number of all abnormal pixel points which are within the minimum reachable range of the abnormal pixel point and belong to the same type of pixel points as the abnormal pixel point, and recording the number as the same type of abnormal pixel points.
3. The stain recognition method for the wallboard mold automatic cleaner according to claim 1, wherein the step of obtaining the density of the abnormal pixel points according to the minimum reachable range and the number of the same kind of abnormal pixel points comprises:
first, the
The calculation formula of the density of each abnormal pixel point is as follows:
in the formula (I), the compound is shown in the specification,
represents a fifth or fifth party>
The density of each abnormal pixel point is judged>
Indicates the fifth->
Number of same type of abnormal pixel points>
Indicates the fifth->
Area of minimum reachable range of abnormal pixel points, based on the area of the abnormal pixel points>
Represents a fifth or fifth party>
The local reachable range radius of each abnormal pixel point.
4. The stain recognition method for the automatic wallboard mold cleaning machine according to claim 1, wherein the step of obtaining the size of the median filter kernel according to all directional noise points of all noise pixel points comprises:
the size of the median filter kernel is calculated as:
in the formula (I), the compound is shown in the specification,
represents the size of the median filtered kernel>
Represents the number of all noisy pixels in all trusted regions, and->
Indicates the fifth->
The noise pixel point and the noise pixel point->
Direction noise point->
The euclidean distance of (c). />
5. The stain recognition method for the automatic wallboard mold cleaning machine according to claim 1, wherein the step of calculating the adhesion degree of all the stain pixel points comprises:
acquiring a circular area which takes the spot pixel points as the center and takes the water spray radius as the radius, counting to acquire the number of all spot pixel points in the circular area, and taking the ratio of the number of all spot pixel points in the circular area to the area of the circular area as the attachment degree of the spot pixel points.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211609040.7A CN115601630B (en) | 2022-12-15 | 2022-12-15 | Stain recognition method for automatic wallboard mold cleaning machine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211609040.7A CN115601630B (en) | 2022-12-15 | 2022-12-15 | Stain recognition method for automatic wallboard mold cleaning machine |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115601630A CN115601630A (en) | 2023-01-13 |
| CN115601630B true CN115601630B (en) | 2023-03-28 |
Family
ID=84853841
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211609040.7A Active CN115601630B (en) | 2022-12-15 | 2022-12-15 | Stain recognition method for automatic wallboard mold cleaning machine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115601630B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115830106B (en) * | 2023-02-16 | 2023-04-25 | 智联信通科技股份有限公司 | Auxiliary positioning method for electrified cleaning of equipment in machine room |
| CN118097310B (en) * | 2024-04-24 | 2024-07-09 | 陕西丰京建材有限公司 | Method for digitally detecting concrete surface defects |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113724243A (en) * | 2021-09-14 | 2021-11-30 | 首都医科大学附属北京天坛医院 | Image processing method, image processing device, electronic equipment and storage medium |
| CN114792329A (en) * | 2021-11-29 | 2022-07-26 | 安徽工业大学 | Color image segmentation method and device based on mesopic clustering and superpixel |
| CN115294338A (en) * | 2022-09-29 | 2022-11-04 | 中威泵业(江苏)有限公司 | Impeller surface defect identification method |
Family Cites Families (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| GB2565775A (en) * | 2017-08-21 | 2019-02-27 | Nokia Technologies Oy | A Method, an apparatus and a computer program product for object detection |
| US10552707B2 (en) * | 2017-12-07 | 2020-02-04 | Qualcomm Incorporated | Methods and devices for image change detection |
| KR20200134813A (en) * | 2019-05-23 | 2020-12-02 | 삼성에스디에스 주식회사 | Apparatus and method for image processing for machine learning |
| CN113052859A (en) * | 2021-04-20 | 2021-06-29 | 哈尔滨理工大学 | Super-pixel segmentation method based on self-adaptive seed point density clustering |
| CN114882468B (en) * | 2022-07-13 | 2022-09-23 | 山东施卫普环保科技有限公司 | Self-adaptive adjustment method for angle of sweeping brush of sweeper |
| CN115147710A (en) * | 2022-07-15 | 2022-10-04 | 杭州电子科技大学 | Sonar image target processing method based on heterogeneous filtering detection and level set segmentation |
| CN115049835B (en) * | 2022-08-16 | 2022-11-29 | 众烁精密模架(南通)有限公司 | Data preprocessing method based on die-casting die defect identification |
| CN115294159B (en) * | 2022-09-30 | 2022-12-23 | 南通市博洋铸造有限公司 | Method for dividing corroded area of metal fastener |
-
2022
- 2022-12-15 CN CN202211609040.7A patent/CN115601630B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113724243A (en) * | 2021-09-14 | 2021-11-30 | 首都医科大学附属北京天坛医院 | Image processing method, image processing device, electronic equipment and storage medium |
| CN114792329A (en) * | 2021-11-29 | 2022-07-26 | 安徽工业大学 | Color image segmentation method and device based on mesopic clustering and superpixel |
| CN115294338A (en) * | 2022-09-29 | 2022-11-04 | 中威泵业(江苏)有限公司 | Impeller surface defect identification method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115601630A (en) | 2023-01-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115601630B (en) | Stain recognition method for automatic wallboard mold cleaning machine | |
| CN110766679B (en) | Lens contamination detection method and device and terminal equipment | |
| US11443437B2 (en) | Vibe-based three-dimensional sonar point cloud image segmentation method | |
| CN109636732B (en) | Hole repairing method of depth image and image processing device | |
| CN107704844B (en) | Power transmission line icing thickness identification method based on binocular parallax images of unmanned aerial vehicle | |
| CN115115612B (en) | Surface defect detection method and system for mechanical parts | |
| CN114723701A (en) | Gear defect detection method and system based on computer vision | |
| CN110517288A (en) | Real-time target detecting and tracking method based on panorama multichannel 4k video image | |
| CN115330767B (en) | Method for identifying production abnormity of corrosion foil | |
| CN104700085B (en) | A kind of chip positioning method based on template matches | |
| CN110414355A (en) | The right bit sky parking stall of view-based access control model and parking stall line detecting method during parking | |
| CN115439494B (en) | Spray image processing method for quality inspection of sprayer | |
| CN114723681A (en) | Concrete crack defect detection method based on machine vision | |
| CN116012357A (en) | Automobile circular trailer hole air line defect detection method based on machine vision | |
| CN115311265B (en) | Loom intelligence control system based on weaving quality | |
| JP2023505663A (en) | Character segmentation method, device and computer readable storage medium | |
| CN115661173B (en) | Remote sensing image-based land parcel segmentation method | |
| CN109378279A (en) | Wafer detection method and wafer detection system | |
| CN115100174B (en) | Ship sheet metal part paint surface defect detection method | |
| CN116703913B (en) | Spraying quality detection method of sprayer | |
| CN114140384A (en) | Transverse vibration image recognition algorithm for hoisting steel wire rope based on contour fitting and centroid tracking | |
| CN112330544A (en) | Image smear processing method, device, equipment and medium | |
| CN115937186A (en) | Textile defect identification method and system | |
| CN116739998A (en) | Coal mine skip roller cage shoe abrasion detection method based on machine vision | |
| CN117557626B (en) | Auxiliary positioning method for spray head installation of aerosol sprayer |
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 |