CN116523909A - Visual detection method and system for appearance of automobile body - Google Patents
Visual detection method and system for appearance of automobile body Download PDFInfo
- Publication number
- CN116523909A CN116523909A CN202310792149.7A CN202310792149A CN116523909A CN 116523909 A CN116523909 A CN 116523909A CN 202310792149 A CN202310792149 A CN 202310792149A CN 116523909 A CN116523909 A CN 116523909A
- Authority
- CN
- China
- Prior art keywords
- image data
- image
- data
- appearance
- extracting
- 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 50
- 230000000007 visual effect Effects 0.000 title claims abstract description 31
- 238000012545 processing Methods 0.000 claims abstract description 44
- 230000007547 defect Effects 0.000 claims abstract description 26
- 230000009467 reduction Effects 0.000 claims abstract description 26
- 238000011179 visual inspection Methods 0.000 claims abstract description 24
- 238000000034 method Methods 0.000 claims abstract description 20
- 238000000605 extraction Methods 0.000 claims abstract description 16
- 238000012216 screening Methods 0.000 claims abstract description 11
- 238000004458 analytical method Methods 0.000 claims description 22
- 238000004364 calculation method Methods 0.000 claims description 22
- 238000001914 filtration Methods 0.000 claims description 16
- 230000009466 transformation Effects 0.000 claims description 9
- 230000002776 aggregation Effects 0.000 claims description 8
- 238000004220 aggregation Methods 0.000 claims description 8
- 230000008859 change Effects 0.000 claims description 8
- 238000005520 cutting process Methods 0.000 claims description 8
- 238000007689 inspection Methods 0.000 claims description 7
- 238000007781 pre-processing Methods 0.000 claims description 7
- 238000011946 reduction process Methods 0.000 claims description 3
- 230000006870 function Effects 0.000 description 7
- 238000005286 illumination Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 3
- 239000003973 paint Substances 0.000 description 3
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000003860 storage Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000005299 abrasion Methods 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/56—Extraction of image or video features relating to colour
-
- 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
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
- G06V10/765—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects using rules for classification or partitioning the feature space
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/08—Detecting or categorising vehicles
Abstract
The invention discloses a method and a system for visual inspection of the appearance of an automobile body, which relate to the technical field of visual inspection, wherein the visual inspection method comprises the following steps: acquiring initial image data of each area of the appearance of the automobile body in the detection area, and performing gray value processing on the initial image data to obtain a preprocessed image data set; comparing the preprocessed image data set with a preset standard image data set, and extracting preprocessed image data with the similarity more than 85% to obtain refined processed data; noise reduction processing is carried out on the refined processing data based on a BM3D algorithm, and image data to be analyzed are obtained; and extracting outline characteristics of the image data to be analyzed, and analyzing the defect condition of the appearance of the automobile based on the standard image data. Image data with poor shooting quality is reduced by screening the preprocessed image data, the system operation load is reduced, the image data is subjected to noise reduction processing by a BM3D algorithm, and the accuracy of feature extraction is improved, so that the accuracy of visual detection is improved.
Description
Technical Field
The invention mainly relates to the technical field of visual detection, in particular to a visual detection method and a visual detection system for the appearance of an automobile body.
Background
In the automobile production process, the automobile body appearance of the automobile product which is required to be assembled is detected, whether the automobile body has defects such as scratches, deformation and abrasion or not is required to be detected, in order to reduce the labor cost and the manual detection error, the existing automobile processing industry adopts a visual detection mode to detect the appearance of the automobile body. The existing visual detection system needs to shoot a plurality of groups of image data for the appearance of the automobile body as analysis data, and the analysis data is integrated to achieve the appearance detection effect of the automobile body.
Disclosure of Invention
The invention aims to overcome the defects of the prior art, and provides a visual detection method and a visual detection system for the appearance of an automobile body.
The invention provides a visual detection method for the appearance of an automobile body, which comprises the following steps:
s11: acquiring initial image data of each area of the appearance of the automobile body in the detection area, and performing gray value processing on the initial image data to obtain a preprocessed image data set;
s12: comparing the preprocessed image data set with a preset standard image data set, and extracting preprocessed image data with the similarity more than 85% to obtain refined processed data;
s13: noise reduction processing is carried out on the refined processing data based on a BM3D algorithm, and image data to be analyzed are obtained;
s14: extracting outline characteristics of the image data to be analyzed, and analyzing the defect condition of the appearance of the automobile body based on the standard image data;
wherein, the step S13 includes:
s131: extracting image data to be detected from the fine processing data, dividing the image to be detected into a plurality of image blocks through a intercepting window with a preset size, and obtaining an image block set;
s132: grouping the image blocks of the image block set into approximate gray values according to a preset gray value change range to obtain a plurality of groups of image block queues;
s133: and calculating Euclidean distances of any two image blocks in each image block queue to obtain array arrangement of Euclidean distances of the image blocks, extracting approximate image block data of each image block, and carrying out collaborative filtering and aggregation.
Further, the obtaining the initial image data of each area of the appearance of the automobile body in the detection area, performing gray value processing on the initial image data, and obtaining the preprocessed image data set includes:
and carrying out graying, geometric transformation and interpolation processing on the initial image data to obtain a preprocessed image data set.
Further, comparing the preprocessed image data set with a preset standard image data set, extracting preprocessed image data with similarity greater than 85%, and obtaining refined processed data includes:
and determining the model of the automobile according to the initial image data of the automobile, and determining corresponding standard image data according to the model of the automobile.
Further, comparing the preprocessed image data set with a preset standard image data set, extracting preprocessed image data with similarity greater than 85%, and obtaining refined processed data further includes:
comparing the preprocessed image data with standard image data of a corresponding area, extracting main features of the preprocessed image and main features of the standard image data, and analyzing similarity of the preprocessed image data and the standard image data through comparison.
Further, the extracting the image data to be detected from the fine processing data, dividing the image to be detected into a plurality of image blocks through a intercepting window with a preset size, and obtaining the image block set includes:
setting the size of the intercepting window asThe preset step length is +.>And driving a cutting window to move in the image data to be detected according to a preset step length until the cutting window traverses all the area of the image data to be detected, and dividing the image data to be detected into a plurality of image blocks.
Further, the calculating the euclidean distance of any two image blocks in each image block queue, and the obtaining the array arrangement of the euclidean distances of the image blocks includes:
taking any two image blocks from the image block queue and calculating the Euclidean distance between the two image blocks, wherein the Euclidean distance has the following calculation formula:
;
wherein ,for the Euclidean distance between two image blocks, < >>For the pixel coordinates of one of the image blocks, for>For the pixel coordinates of another image block, k is a constant and n is the number of pixels in the image block.
Further, the calculating the euclidean distance of any two image blocks in each image block queue, obtaining an array arrangement of the euclidean distances of the image blocks, and extracting the approximate image block data of each image block for collaborative filtering and aggregation includes:
performing Euclidean distance calculation on the image blocks of the image block queue in an array arrangement mode to construct an array arrangement coordinate system;
extracting image block data in the array arrangement coordinate system and the Euclidean distance of the image block dataAnd obtaining a three-dimensional array of similar image blocks.
Further, the extracting the image block data in the array arrangement coordinate system and the image block data with the Euclidean distance adjacent to each otherThe image blocks include:
setting a judgment threshold value of the Euclidean distanceExtracting an image block from the image block queue as a basic block, inquiring Euclidean distance calculation data with the basic block in the array arrangement coordinate system, and comparing the obtained Euclidean distance calculation data with the judgment threshold value->Comparing, if the Euclidean distance calculation data is less than or equal to the judgment threshold +.>And extracting the image block of the corresponding Euclidean distance calculation data as a similar image block of the basic block.
Further, the extracting the outline features of the image data to be analyzed, and analyzing the defect condition of the appearance of the automobile body based on the standard image data includes:
and performing contour feature extraction on the image data to be analyzed through ORB feature extraction, and performing contour feature comparison on the analysis image by combining the contour features of the standard image data, so as to detect the defect condition of the appearance of the automobile body.
The invention also provides a visual detection system for the appearance of the automobile body, which comprises a plurality of visual detection parts and a main control system, wherein the main control system is electrically connected with the plurality of visual detection parts;
the visual detection components are used for acquiring initial image data of each area of the appearance of the automobile body in the detection area and sending the initial image data to the main control system;
the main control system is internally provided with a preprocessing module, a screening module, a noise reduction module and a defect analysis module;
the preprocessing module is used for performing gray value processing on the initial image data to obtain a preprocessed image data set;
the screening module is used for comparing the preprocessed image data set with a preset standard image data set, extracting preprocessed image data with the similarity being more than 85%, and obtaining refined processed data;
the noise reduction module is used for carrying out noise reduction processing on the refined processing data through a BM3D algorithm to obtain image data to be analyzed;
the noise reduction process includes:
extracting image data to be detected from the fine processing data, dividing the image to be detected into a plurality of image blocks through a intercepting window with a preset size, and obtaining an image block set;
grouping the image blocks of the image block set into approximate gray values according to a preset gray value change range to obtain a plurality of groups of image block queues;
calculating Euclidean distance of any two image blocks in each image block queue, obtaining array arrangement of Euclidean distance of the image blocks, extracting approximate image block data of each image block, and carrying out collaborative filtering and aggregation;
the defect analysis module is used for extracting outline characteristics of the image data to be analyzed and analyzing defect conditions of the appearance of the automobile body based on the standard image data.
The invention provides a visual detection method and a visual detection system for the appearance of an automobile body, which are used for reducing the interference of image data with poor shooting quality by screening preprocessed image data, reducing the operation load of a detection system, and improving the accuracy of feature extraction by performing noise reduction processing on the image data through a BM3D algorithm, thereby improving the accuracy of visual detection.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions in the prior art, the drawings which are required in the description of the embodiments or the prior art will be briefly described, it being obvious that the drawings in the description below are only some embodiments of the invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic flow chart of a visual inspection method for the appearance of an automobile body in an embodiment of the invention;
fig. 2 is a schematic diagram of a noise reduction processing flow for the refined data based on a BM3D algorithm in an embodiment of the present invention;
FIG. 3 is a schematic diagram of a visual inspection system for the appearance of an automotive body in accordance with an embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Fig. 1 shows a schematic flow chart of a visual inspection method for automobile body appearance according to an embodiment of the invention, where the visual inspection method includes:
s11: and acquiring initial image data of each area of the appearance of the automobile body in the detection area, and performing gray value processing on the initial image data to obtain a preprocessed image data set.
Shooting all areas of the appearance of the automobile body in the detection area through a plurality of visual detection parts to obtain initial image data, wherein each visual detection part is correspondingly arranged on each area of the appearance of the automobile body, and any visual detection part continuously shoots a certain area of the appearance of the automobile body, so that enough initial image data are obtained, and the accuracy of the appearance detection of the automobile body is improved.
Specifically, the performing gray value processing on the initial image data to obtain a preprocessed image data set includes:
and the main control system performs graying, geometric transformation and interpolation processing on the initial image data to obtain a preprocessed image data set.
Specifically, the gray level of the initial image data is converted into gray level of the initial image data to output new image data, and the geometric transformation can change the spatial relationship between objects in the image data. Common modes of operation for geometric operations are affine transformation (Affine Transformation) and image wrapping (ImageWarping).
Further, the interpolation processing is gray level interpolation, in this embodiment, gray level interpolation is performed by a nearest neighbor interpolation method, that is, gray values of neighboring points of the mapping positions corresponding to the pixel points of the initial image data are found to be used as interpolation results, so that the method has the characteristics of simplicity, rapidness and good gray level fidelity, can meet the requirement of preliminary processing of the appearance image of the automobile body, reduces the computational complexity, and improves the visual detection efficiency of the appearance of the automobile body.
Specifically, by preprocessing the initial image data, irrelevant information in the initial image data can be reduced, and relevant information of the appearance of the automobile body can be enhanced, so that the accuracy of visual detection of the appearance of the automobile body can be improved.
S12: and comparing the preprocessed image data set with a preset standard image data set, and extracting preprocessed image data with the similarity more than 85% to obtain refined processed data.
Specifically, the model of the automobile is determined according to the initial image data of the automobile, corresponding standard image data is determined according to the model of the automobile, the main control system obtains the initial image data, the model and the color of the automobile are identified according to the initial image data, and the automobile body appearance image data corresponding to the model and the color is extracted from an automobile body appearance standard image database based on the model and the color of the automobile, so that the standard image data is obtained.
Furthermore, the automobile body appearance standard image database is constructed according to actual detection requirements, and automobile body appearance standard image data of all types and all colors of automobiles to be detected are recorded in the automobile body appearance standard image database, so that corresponding automobile body appearance standard image data can be found according to the types of detected automobiles.
Comparing the preprocessed image data with standard image data of a corresponding area, extracting main features of the preprocessed image and main features of the standard image data, and analyzing similarity of the preprocessed image data and the standard image data through comparison.
Specifically, the appearance condition of the automobile body of the preprocessed image data is judged according to the similarity between the preprocessed image data and the standard image data, when the similarity between the preprocessed image data and the standard image data is more than 85%, the preprocessed image data is judged to reflect the real image of the appearance of the automobile body, and the preprocessed image data is reserved to obtain the refined processed data.
Further, if the similarity between the preprocessed image data and the standard image data is less than 85%, the preprocessed image data is judged to be greatly interfered, and the main control system marks the preprocessed image data and sorts the preprocessed image data into defect image data.
Further, after the appearance detection of the automobile body is finished, the main control system re-acquires the image data of the area corresponding to the defect image data according to the defect image data to carry out re-detection, so that the reliability and the accuracy of the appearance detection of the automobile body are improved.
S13: and carrying out noise reduction treatment on the refined processing data based on a BM3D algorithm to obtain image data to be analyzed.
Specifically, a BM3D (Block Matching 3D) algorithm is a three-dimensional Matching algorithm, and an image is divided into a plurality of image blocks, adjacent image blocks are matched to form a three-dimensional matrix, the three-dimensional matrix is filtered in a three-dimensional space, and the result is inversely transformed and fused to two dimensions to form a denoised image.
Fig. 2 shows a schematic flow chart of noise reduction processing for the refined data based on the BM3D algorithm in an embodiment of the present invention, where the step S13 includes:
s131: extracting image data to be detected from the fine processing data, and dividing the image to be detected into a plurality of image blocks through a intercepting window with a preset size to obtain an image block set.
Extracting image data to be detected from the fine processing data, and setting the size of a cutting window to be the same as the preset size by setting the cutting window to be the preset sizeThe preset step length is +.>And driving a cutting window to move in the image data to be detected according to a preset step length until the cutting window traverses all the area of the image data to be detected, and dividing the image data to be detected into a plurality of image blocks.
Further, the preset step lengthThe constraint relation with the interception window is as follows: />；/>；
Ensuring that the intercepting window can completely traverse the image data to be detected, namely ensuring that a plurality of image blocks comprise all image data contents of the image data to be detected.
S132: and grouping the image blocks of the image block set into approximate gray values according to a preset gray value change range to obtain a plurality of groups of image block queues.
Specifically, according to the gray value of each image block, dividing a plurality of image blocks with approximate gray values into the same queues, so as to obtain a plurality of groups of image block queues, and dividing the image blocks with approximate gray values into the same queues, so as to improve the searching convenience of the approximate image blocks, reduce the calculated amount of a BM3D algorithm, and improve the noise reduction efficiency.
S133: and calculating Euclidean distances of any two image blocks in each image block queue to obtain array arrangement of Euclidean distances of the image blocks, extracting approximate image block data of each image block, and carrying out collaborative filtering and aggregation.
Specifically, any two image blocks are taken from the image block queue, and the Euclidean distance between the two image blocks is calculated, wherein the Euclidean distance is calculated according to the formula:
;
wherein ,for the Euclidean distance between two image blocks, < >>For the pixel coordinates of one of the image blocks, for>For the pixel coordinates of another image block, k is a constant and n is the number of pixels in the image block.
Specifically, performing pairwise Euclidean distance calculation on the image blocks in the image block queue in an array arrangement mode, arranging the image blocks along the X-axis and Y-axis directions according to the number of the image blocks in the image block queue to form an array arrangement coordinate system, performing pairwise Euclidean distance calculation on the image blocks in the image block queue, and filling calculation results into the array arrangement coordinate system until all data nodes in the array arrangement coordinate system are filled, thereby finishing Euclidean distance calculation of all the image blocks.
Further, by extracting the image block data in the array arrangement and the image block data with Euclidean distance adjacent to each otherAnd obtaining the three-dimensional array of similar image blocks.
Further, the saidAnd for the preset analysis quantity of the similar image blocks, the analysis quantity can be adjusted according to the actual detection requirement, so that the accuracy of the visual detection result is improved.
Specifically, a judgment threshold value of the Euclidean distance is setExtracting an image block from the image block queue as a basic block, inquiring Euclidean distance calculation data with the basic block in the array arrangement coordinate system, and comparing the obtained Euclidean distance calculation data with the judgment threshold value->Comparing, if the Euclidean distance calculation data is less than or equal to the judgment threshold +.>Extracting image blocks of corresponding Euclidean distance calculation data as similar image blocks of the basic blocks, and when the number of the obtained similar image blocks reaches the preset analysis number of the similar image blocks +.>Based on the->And the similar image blocks form a three-dimensional array for the basic blocks.
Further, analysis of similar image blocks is sequentially carried out on each image block, and a three-dimensional array of each image block is obtained.
Specifically, each three-dimensional array is subjected to Fourier transform, and the three-dimensional arrays after Fourier transform are filtered, so that the filtered and noise-reduced image block is obtained.
The fourier transform formula is:
;
;
wherein ,is three-dimensional array->Is->Is a two-dimensional expansion function of>For the length of each dimension in the three-dimensional array, +.>Is complex unit root, i is imaginary unit.
For functions obtained after Fourier transformationGaussian filtering is carried out, data after filtering is obtained through neighborhood weighted average calculation on each data point in the function, and the function is subjected to +.Pv2.Gaussian Blur () function is called to perform +.>And carrying out Gaussian filtering to obtain noise reduction data after two-dimensional transformation.
Furthermore, the cv2.Gaussian Blur () function is a Gaussian blur function, and is used for realizing Gaussian filtering, and performing linear smoothing filtering on data to achieve the denoising effect.
Further, the fourier transform further includes performing one-dimensional transform on third dimensional data of the three-dimensional array, extracting a change in a time dimension of the three-dimensional array, performing mean filtering on the transformed one-dimensional data, dividing the transformed one-dimensional data into a plurality of data blocks, performing mean filtering on each block, and finally integrating to obtain noise reduction data after the one-dimensional data filtering.
Further, the noise reduction data of the one-dimensional data are combined with the noise reduction data after the two-dimensional transformation of the three-dimensional array, and the noise reduction array data of the three-dimensional array are obtained through integration.
Specifically, the noise reduction array data is subjected to three-dimensional inverse transformation to obtain an estimated value of each image block, and the estimated value of each image block is integrated to obtain image data to be analyzed.
S14: and extracting contour features of the image data to be analyzed, and analyzing the defect condition of the appearance of the automobile body based on the standard image data.
Specifically, contour feature extraction is performed on the image data to be analyzed through ORB feature extraction, contour feature comparison is performed on the analysis image by combining the contour features of the standard image data, and therefore the defect condition of the appearance of the automobile body is detected.
Specifically, the ORB (Oriented FAST and Rotated BRIEF) algorithm is to extract feature points based on a FAST (Features from Accelerated Segment Test) acceleration segment test feature algorithm, construct descriptors of the feature points based on the BRIEF (Binary Robust Independent Elementary Features) algorithm, and realize scale invariance and rotation invariance of the feature points, namely the feature points after scaling and rotation can still generate descriptors similar to the original descriptors.
Specifically, the ORB algorithm extracts a large number of feature points from the analysis image data by performing feature rough extraction on the analysis image data, searches a search circle with a preset radius for comparing gray values of a comparison point and a judgment point in the search circle, and considers the judgment point as a feature point if a difference exists between the gray value of a continuous pixel point and the gray value of the judgment point in the search circle.
Further, in this embodiment, the radius of the search circle is 20 pixels, the number of the continuous pixels is set to 10, and the comparison points of the odd number pixel distances from the judgment point are preferentially detected, so that the search efficiency is improved.
Further, the obtained characteristic points are screened in a machine learning mode, the outline of the automobile in the analysis image is identified through a pre-trained screening model, the whole area of the automobile in the analysis image data is obtained, the characteristic points in the whole area of the automobile are reserved, the characteristic points outside the whole area of the automobile are extracted, and interference of the characteristic points is reduced.
Further, linear convergence is carried out on the local dense feature points, corresponding direction routing is set according to the overall shape of the local dense feature points, distance deviation of each feature point from the direction routing is calculated, position correction is carried out on the feature points according to the distance deviation, and each pixel point is close to the direction routing to be converged, so that feature extraction accuracy is improved.
Further, the direction of the feature point is determined by moment (movement), that is, the centroid is obtained by moment calculation in the range of the preset radius by taking the feature point as the center of a circle, and the vector is formed from the feature point to the centroid and is set as the direction of the feature point.
Furthermore, the descriptor is constructed through BRIEF algorithm, the region of the descriptor is determined, the establishment speed of the descriptor is improved through a random response party, and the FAST algorithm is combined to achieve convenient feature extraction.
Specifically, feature contrast analysis is performed on the image data to be analyzed with the feature extraction completed and the standard image data, and feature points, which are different from the standard image data, in the image data to be analyzed are marked, so that the defect condition in the image area to be analyzed is determined.
Further, the defect conditions of the appearance of the automobile body comprise scratches, paint dropping, deformation and the like, and the defect position marking of the image data of the automobile body can be marked through characteristic analysis.
The embodiment of the invention provides a visual detection method for the appearance of an automobile body, which reduces the interference of image data with poor shooting quality by screening preprocessed image data, reduces the operation load of a detection system, and improves the accuracy of feature extraction by performing noise reduction processing on the image data through a BM3D algorithm, thereby improving the accuracy of visual detection.
Embodiment two:
fig. 3 shows a schematic structural diagram of a visual inspection system for the appearance of an automobile body in an embodiment of the present invention, where the system includes a main control system 1 and a plurality of visual inspection units 2, the plurality of visual inspection units 2 are electrically connected to the main control system 1, and the plurality of visual inspection units 2 are configured to acquire initial image data of each area of the appearance of the automobile body in an inspection area, and send the initial image data to the main control system 1;
a preprocessing module 11, a screening module 12, a noise reduction module 13 and a defect analysis module 14 are arranged in the main control system 1;
the preprocessing module 11 is configured to perform gray value processing on the initial image data to obtain a preprocessed image data set;
the screening module 12 is configured to compare the preprocessed image data set with a preset standard image data set, and extract preprocessed image data with a similarity greater than 85% to obtain refined processed data;
the noise reduction module 13 is configured to perform noise reduction processing on the refined processing data through a BM3D algorithm to obtain image data to be analyzed;
the noise reduction process includes:
extracting image data to be detected from the fine processing data, dividing the image to be detected into a plurality of image blocks through a intercepting window with a preset size, and obtaining an image block set;
grouping the image blocks of the image block set into approximate gray values according to a preset gray value change range to obtain a plurality of groups of image block queues;
calculating Euclidean distance of any two image blocks in each image block queue, obtaining array arrangement of Euclidean distance of the image blocks, extracting approximate image block data of each image block, and carrying out collaborative filtering and aggregation;
the defect analysis module 14 is configured to perform contour feature extraction on the image data to be analyzed, and analyze a defect condition of an appearance of an automobile body based on the standard image data.
Specifically, the vision inspection system is further provided with an illumination component 3, the illumination component 3 is correspondingly arranged on the side edges of the vision inspection components 2, and the illumination component 3 can provide enough illumination to ensure that the vision inspection components 2 can meet the vision inspection requirements of the appearance of the automobile body.
Furthermore, the illumination component 3 can adopt a diffuse reflection annular light source, is suitable for visual detection of the paint surface of the automobile body, and reduces the influence of illumination reflection of the paint surface of the automobile body, thereby improving the reliability of appearance detection of the automobile body.
The embodiment of the invention provides a system for visual inspection of the appearance of an automobile body, which is used for reducing the interference of image data with poor shooting quality by screening preprocessed image data, reducing the operation load of the inspection system, and improving the accuracy of feature extraction by performing noise reduction processing on the image data through a BM3D algorithm, thereby improving the accuracy of visual inspection.
Those of ordinary skill in the art will appreciate that all or part of the steps in the various methods of the above embodiments may be implemented by a program to instruct related hardware, the program may be stored in a computer readable storage medium, and the storage medium may include: read Only Memory (ROM), random access Memory (RAM, random Access Memory), magnetic or optical disk, and the like.
In addition, the visual inspection method and system for the appearance of the automobile body provided by the embodiment of the invention are described in detail, and specific examples are adopted to illustrate the principle and implementation of the invention, and the description of the above examples is only used for helping to understand the method and core idea of the invention; meanwhile, as those skilled in the art will have variations in the specific embodiments and application scope in accordance with the ideas of the present invention, the present description should not be construed as limiting the present invention in view of the above.
Claims (10)
1. A visual inspection method for the appearance of an automotive body, the visual inspection method comprising:
s11: acquiring initial image data of each area of the appearance of the automobile body in the detection area, and performing gray value processing on the initial image data to obtain a preprocessed image data set;
s12: comparing the preprocessed image data set with a preset standard image data set, and extracting preprocessed image data with the similarity more than 85% to obtain refined processed data;
s13: noise reduction processing is carried out on the refined processing data based on a BM3D algorithm, and image data to be analyzed are obtained;
s14: extracting outline characteristics of the image data to be analyzed, and analyzing the defect condition of the appearance of the automobile body based on the standard image data;
wherein, the step S13 includes:
s131: extracting image data to be detected from the fine processing data, dividing the image to be detected into a plurality of image blocks through a intercepting window with a preset size, and obtaining an image block set;
s132: grouping the image blocks of the image block set into approximate gray values according to a preset gray value change range to obtain a plurality of groups of image block queues;
s133: and calculating Euclidean distances of any two image blocks in each image block queue to obtain array arrangement of Euclidean distances of the image blocks, extracting approximate image block data of each image block, and carrying out collaborative filtering and aggregation.
2. The visual inspection method for the appearance of an automotive body according to claim 1, wherein the acquiring the initial image data of each area of the appearance of the automotive body in the inspection area, performing gray-value processing on the initial image data, and obtaining the preprocessed image data group includes:
and carrying out graying, geometric transformation and interpolation processing on the initial image data to obtain a preprocessed image data set.
3. The visual inspection method for the appearance of an automotive body according to claim 1, wherein comparing the preprocessed image data set with a preset standard image data set, extracting preprocessed image data having a similarity of more than 85%, and obtaining refined processed data comprises:
and determining the model of the automobile according to the initial image data of the automobile, and determining corresponding standard image data according to the model of the automobile.
4. The visual inspection method for the appearance of an automotive body according to claim 1, wherein comparing the preprocessed image data set with a preset standard image data set, extracting preprocessed image data having a similarity of more than 85%, and obtaining refined processed data further comprises:
comparing the preprocessed image data with standard image data of a corresponding area, extracting main features of the preprocessed image and main features of the standard image data, and analyzing similarity of the preprocessed image data and the standard image data through comparison.
5. The visual inspection method for the appearance of an automobile body according to claim 1, wherein the extracting image data to be inspected from the refined processing data, dividing the image to be inspected into a plurality of image blocks through an intercepting window with a preset size, and obtaining an image block set includes:
setting the size of the intercepting window asThe preset step length is +.>And driving a cutting window to move in the image data to be detected according to a preset step length until the cutting window traverses all the area of the image data to be detected, and dividing the image data to be detected into a plurality of image blocks.
6. The visual inspection method for the appearance of an automobile body according to claim 1, wherein the calculating the euclidean distance of any two image blocks in each image block queue to obtain the array arrangement of the euclidean distances of the image blocks comprises:
taking any two image blocks from the image block queue and calculating the Euclidean distance between the two image blocks, wherein the Euclidean distance has the following calculation formula:
;
wherein ,for the Euclidean distance between two image blocks, < >>For the pixel coordinates of one of the image blocks, for>For the pixel coordinates of another image block, k is a constant and n is the number of pixels in the image block.
7. The visual inspection method for the appearance of an automobile body according to claim 1, wherein the calculating the euclidean distance of any two image blocks in each image block queue to obtain an array arrangement of the euclidean distances of the image blocks, and the extracting the approximate image block data of each image block for collaborative filtering and aggregation comprises:
performing Euclidean distance calculation on the image blocks of the image block queue in an array arrangement mode to construct an array arrangement coordinate system;
extracting image block data in the array arrangement coordinate system and the Euclidean distance of the image block dataAnd obtaining a three-dimensional array of similar image blocks.
8. The visual inspection method for the appearance of an automotive body according to claim 7, wherein said extracting image block data in said array arrangement coordinate system and said image block data being euclidean distance-adjacentThe image blocks include:
setting a judgment threshold value of the Euclidean distanceExtracting an image block from the image block queue as a basic block, inquiring Euclidean distance calculation data with the basic block in the array arrangement coordinate system, and comparing the obtained Euclidean distance calculation data with the judgment threshold value->Comparing, if the Euclidean distance calculation data is less than or equal to the judgment threshold +.>And extracting the image block of the corresponding Euclidean distance calculation data as a similar image block of the basic block.
9. The visual inspection method for the appearance of an automobile body according to claim 1, wherein the contour feature extraction of the image data to be analyzed and the analysis of the defect condition of the appearance of the automobile body based on the standard image data comprise:
and performing contour feature extraction on the image data to be analyzed through ORB feature extraction, and performing contour feature comparison on the analysis image by combining the contour features of the standard image data, so as to detect the defect condition of the appearance of the automobile body.
10. The visual detection system for the appearance of the automobile body is characterized by comprising a plurality of visual detection components and a master control system, wherein the master control system is electrically connected with the plurality of visual detection components;
the visual detection components are used for acquiring initial image data of each area of the appearance of the automobile body in the detection area and sending the initial image data to the main control system;
the main control system is internally provided with a preprocessing module, a screening module, a noise reduction module and a defect analysis module;
the preprocessing module is used for performing gray value processing on the initial image data to obtain a preprocessed image data set;
the screening module is used for comparing the preprocessed image data set with a preset standard image data set, extracting preprocessed image data with the similarity being more than 85%, and obtaining refined processed data;
the noise reduction module is used for carrying out noise reduction processing on the refined processing data through a BM3D algorithm to obtain image data to be analyzed;
the noise reduction process includes:
extracting image data to be detected from the fine processing data, dividing the image to be detected into a plurality of image blocks through a intercepting window with a preset size, and obtaining an image block set;
grouping the image blocks of the image block set into approximate gray values according to a preset gray value change range to obtain a plurality of groups of image block queues;
calculating Euclidean distance of any two image blocks in each image block queue, obtaining array arrangement of Euclidean distance of the image blocks, extracting approximate image block data of each image block, and carrying out collaborative filtering and aggregation;
the defect analysis module is used for extracting outline characteristics of the image data to be analyzed and analyzing defect conditions of the appearance of the automobile body based on the standard image data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310792149.7A CN116523909B (en) | 2023-06-30 | 2023-06-30 | Visual detection method and system for appearance of automobile body |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310792149.7A CN116523909B (en) | 2023-06-30 | 2023-06-30 | Visual detection method and system for appearance of automobile body |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116523909A true CN116523909A (en) | 2023-08-01 |
| CN116523909B CN116523909B (en) | 2023-09-29 |
Family
ID=87399752
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310792149.7A Active CN116523909B (en) | 2023-06-30 | 2023-06-30 | Visual detection method and system for appearance of automobile body |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116523909B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108491813A (en) * | 2018-03-29 | 2018-09-04 | 百度在线网络技术（北京）有限公司 | Method and apparatus for fresh information |
| CN109781739A (en) * | 2019-03-04 | 2019-05-21 | 杭州晶耐科光电技术有限公司 | Automobile finish surface appearance defects automatic detection system and method |
| CN110879230A (en) * | 2019-12-13 | 2020-03-13 | 江苏海洋大学 | Intelligent detection system for automobile defects |
| CN116188475A (en) * | 2023-05-05 | 2023-05-30 | 德中(深圳)激光智能科技有限公司 | Intelligent control method, system and medium for automatic optical detection of appearance defects |
-
2023
- 2023-06-30 CN CN202310792149.7A patent/CN116523909B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108491813A (en) * | 2018-03-29 | 2018-09-04 | 百度在线网络技术（北京）有限公司 | Method and apparatus for fresh information |
| CN109781739A (en) * | 2019-03-04 | 2019-05-21 | 杭州晶耐科光电技术有限公司 | Automobile finish surface appearance defects automatic detection system and method |
| CN110879230A (en) * | 2019-12-13 | 2020-03-13 | 江苏海洋大学 | Intelligent detection system for automobile defects |
| CN116188475A (en) * | 2023-05-05 | 2023-05-30 | 德中(深圳)激光智能科技有限公司 | Intelligent control method, system and medium for automatic optical detection of appearance defects |
Non-Patent Citations (1)
| Title |
|---|
| 郑文斌 等: "汽车玻璃黑边区域划痕的视觉检测技术", 福建工程学院学报, no. 01 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116523909B (en) | 2023-09-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113450307B (en) | Product edge defect detection method | |
| CN108898610B (en) | Object contour extraction method based on mask-RCNN | |
| CN108520274B (en) | High-reflectivity surface defect detection method based on image processing and neural network classification | |
| CN110866924B (en) | Line structured light center line extraction method and storage medium | |
| JP2940317B2 (en) | Image processing device | |
| CN109580630B (en) | Visual inspection method for defects of mechanical parts | |
| CN107230203B (en) | Casting defect identification method based on human eye visual attention mechanism | |
| CN111723721A (en) | Three-dimensional target detection method, system and device based on RGB-D | |
| CN111507390A (en) | Storage box body identification and positioning method based on contour features | |
| CN112613378B (en) | 3D target detection method, system, medium and terminal | |
| CN104299260A (en) | Contact network three-dimensional reconstruction method based on SIFT and LBP point cloud registration | |
| CN110021029B (en) | Real-time dynamic registration method and storage medium suitable for RGBD-SLAM | |
| CN115797813B (en) | Water environment pollution detection method based on aerial image | |
| CN115018846A (en) | AI intelligent camera-based multi-target crack defect detection method and device | |
| CN111783722B (en) | Lane line extraction method of laser point cloud and electronic equipment | |
| CN111091071B (en) | Underground target detection method and system based on ground penetrating radar hyperbolic wave fitting | |
| CN112257721A (en) | Image target region matching method based on Fast ICP | |
| CN110648359B (en) | Fruit target positioning and identifying method and system | |
| CN113688846A (en) | Object size recognition method, readable storage medium, and object size recognition system | |
| CN107170004B (en) | Image matching method for matching matrix in unmanned vehicle monocular vision positioning | |
| CN116523909B (en) | Visual detection method and system for appearance of automobile body | |
| CN110717910B (en) | CT image target detection method based on convolutional neural network and CT scanner | |
| CN116958837A (en) | Municipal facilities fault detection system based on unmanned aerial vehicle | |
| CN115034577A (en) | Electromechanical product neglected loading detection method based on virtual-real edge matching | |
| CN113505811A (en) | Machine vision imaging method for hub production |
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 |