CN107607546B - Leather defect detection method, system and device based on photometric stereo vision - Google Patents
Leather defect detection method, system and device based on photometric stereo vision Download PDFInfo
- Publication number
- CN107607546B CN107607546B CN201710847126.6A CN201710847126A CN107607546B CN 107607546 B CN107607546 B CN 107607546B CN 201710847126 A CN201710847126 A CN 201710847126A CN 107607546 B CN107607546 B CN 107607546B
- Authority
- CN
- China
- Prior art keywords
- leather
- image
- images
- curvature
- gray
- 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
Images
Landscapes
- Image Processing (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
- Length Measuring Devices By Optical Means (AREA)
Abstract
The invention discloses a leather defect detection method, a leather defect detection system and a leather defect detection device based on photometric stereo vision. The method comprises the steps of collecting a plurality of first images shot under the action of light sources with different positions; and synthesizing the plurality of first images into a depth image, sequentially performing curvature filtering, gray scale stretching and binarization processing to obtain a binary image of the leather to be detected, processing the binary image by using an SVM (support vector machine), and outputting a leather defect detection result. The device comprises a memory and a processor for executing the detection method. By using the method and the device, the enhancement effect of the leather defect area can be rapidly improved, and the leather defect detection accuracy and the treatment efficiency are improved. The leather defect detection method, the system and the device based on the photometric stereo vision can be widely applied to the field of leather detection.
Description
Technical Field
The invention relates to a defect detection technology, in particular to a leather defect detection method, a leather defect detection system and a leather defect detection device based on photometric stereo vision.
Background
Explanation of technical words:
photometric Stereo: and (4) photometric stereo.
SVM: referred to as support vector machines.
Quality monitoring is particularly important in the production process of leather products such as leather bags, automobile seats, clothes and the like, and therefore, the defect detection and positioning of leather as a raw material of the leather products are important work of quality monitoring. At present, the defect detection of leather in the industry mainly depends on human eyes, however, the noisy factory environment, a large amount of detection work and the complexity of leather defect types make the quality control requirements difficult to meet by a detection mode only depending on human eyes. Therefore, in order to solve these problems, it has been proposed to perform defect detection of leather using machine vision techniques.
The machine vision technology has the advantages of improving the production flexibility and the automation degree, so that the defect detection of the leather is realized by utilizing the machine vision technology, and the machine vision technology can be applied to dangerous working environments which are not suitable for manual operation or occasions where the quality control requirements are difficult to meet by manual vision, and has high applicability; meanwhile, in the process of mass industrial production, the defect detection of the leather is realized by utilizing a machine vision technology, so that the production efficiency and the automation degree of production can be greatly improved, and the detection accuracy can be improved. However, the leather defect detection technology based on machine vision, which is commonly used at present, still has some disadvantages, such as low defect identification and detection precision, low running processing efficiency, and the like.
Disclosure of Invention
In order to solve the above technical problems, a first object of the present invention is to provide a leather defect detecting method based on photometric stereo, which can improve the defect detecting accuracy and processing efficiency of the leather defect detecting technology.
The second purpose of the invention is to provide a leather defect detection system based on photometric stereo, which can improve the defect detection accuracy and processing efficiency of the leather defect detection technology.
The third purpose of the invention is to provide a leather defect detection device based on photometric stereo vision, which can improve the defect detection precision and the processing efficiency of the leather defect detection technology.
The first technical scheme adopted by the invention is as follows: the leather defect detection method based on photometric stereo vision comprises the following steps:
acquiring a plurality of first images shot under the irradiation action of light sources at different positions, wherein the first images refer to images of leather to be detected;
synthesizing a plurality of acquired first images by adopting a photometric stereo vision algorithm to obtain a second image;
curvature filtering is carried out on the second image to obtain a curvature map;
performing gray stretching processing on the curvature map to obtain a gray stretched image;
carrying out image binarization processing on the gray level stretching image to obtain a binary image of the leather to be detected;
and processing the binary image of the leather to be detected by using a support vector machine, and outputting the detection result of the leather to be detected.
Further, the step of acquiring a plurality of first images captured under the irradiation of light sources at different positions includes the following steps:
fixing the position of the camera, and enabling the camera to be perpendicular to the leather detection platform;
arranging a plurality of light sources at different positions above the leather detection platform, and enabling an included angle between incident light of the light sources and the leather detection platform to be 30-60 degrees;
controlling a plurality of light sources to generate different illumination conditions;
controlling the camera to take a first image under a lighting condition;
and acquiring a first image obtained by shooting.
Further, the step of synthesizing the plurality of acquired first images into a second image by using a photometric stereo algorithm includes the following steps:
calculating a surface normal vector of the leather to be measured according to the plurality of collected first images;
calculating an orthogonal vector corresponding to the surface normal vector according to the calculated surface normal vector;
and generating a depth image according to the calculated orthogonal vector.
Further, the step of obtaining a curvature map after curvature filtering is performed on the second image specifically includes: and performing curvature filtering on the second image by adopting a Gaussian curvature filter to obtain a curvature map.
Further, the expression formula of the gaussian curvature filter is as follows:
where U represents the second image of the input, H represents the curvature map of the output, and x represents the image convolution operation.
Further, the step of performing gray scale stretching processing on the curvature map comprises the following steps:
obtaining the minimum gray value and the maximum gray value of the curvature map;
and stretching the gray value of the pixel point in the curvature map by using the minimum gray value and the maximum gray value.
Further, the step of performing gray stretching on the gray value of the pixel point in the curvature map by using the minimum gray value and the maximum gray value is as follows:
wherein Y represents a gray value after gray stretching, X represents a gray value before gray stretching, and X representsminExpressed as the minimum gray value, XmaxExpressed as the maximum gray value.
Further, in the step of performing image binarization processing on the gray-scale stretched image, a binarization conversion formula adopted by the step is as follows:
threshold＝E(S(x))×3
where o (x) is expressed as an output binary image, s (x) is expressed as a pixel value of a pixel point in the grayscale stretched image, threshold is expressed as a threshold, and E (s (x)) is expressed as an expected value of the grayscale stretched image.
The second technical scheme adopted by the invention is as follows: leather defect detecting system based on luminosity stereovision, this system includes:
the first acquisition module is used for acquiring a plurality of first images shot under the irradiation action of light sources at different positions, wherein the first images refer to images of the leather to be detected;
the first processing module is used for synthesizing a plurality of acquired first images to obtain a second image by adopting a photometric stereo vision algorithm;
the second processing module is used for performing curvature filtering on the second image to obtain a curvature map;
the third processing module is used for performing gray stretching processing on the curvature map to obtain a gray stretching image;
the fourth processing module is used for carrying out image binarization processing on the gray level stretching image to obtain a binary image of the leather to be detected;
and the fifth processing module is used for processing the binary image of the leather to be detected by using the support vector machine and outputting the detection result of the leather to be detected.
The third technical scheme adopted by the invention is as follows: leather defect detection device based on luminosity stereovision, the device includes:
a memory for storing at least one program;
and the processor is used for loading the at least one program and executing the leather defect detection method based on the photometric stereo vision.
The method, the system and the device have the advantages that: the invention collects a plurality of images of the leather to be measured under the irradiation of light sources at different positions, then synthesizes the collected images by adopting a photometric stereo vision algorithm to obtain a second image with certain three-dimensional information, sequentially carrying out curvature filtering processing, gray scale stretching processing and binarization processing on the second image to obtain a binary image of the leather to be detected, finally processing the binary image of the leather to be detected by using a trained support vector machine, outputting a defect detection result of the leather to be detected, therefore, by using the technology of the invention, the enhancement effect of the leather defect area in the leather image to be detected can be rapidly improved, therefore, the leather defect detection precision and the work processing efficiency are greatly improved, the workload of workers is reduced, and the manufacturing quality and efficiency of leather finished products can be improved.
Drawings
FIG. 1 is a flow chart of the steps of a leather defect detection method based on photometric stereo vision according to the present invention;
FIG. 2 is a block diagram of the leather defect detection system based on photometric stereo vision according to the present invention;
FIG. 3 is a flowchart illustrating steps of an embodiment of SVM construction in the leather defect detection method based on photometric stereo vision according to the present invention;
FIG. 4 is a schematic structural diagram of a photometric stereo experiment platform;
FIG. 5 is a schematic diagram of the comparison of the images of the leather to be detected, the synthetic image and the detection result under different illumination conditions.
Detailed Description
Example 1
As shown in fig. 1, the leather defect detection method based on photometric stereo vision comprises the following steps:
acquiring a plurality of first images shot under the irradiation action of light sources at different positions, wherein the first images refer to images of leather to be detected;
synthesizing a plurality of acquired first images by adopting a photometric stereo vision algorithm to obtain a second image;
curvature filtering is carried out on the second image to obtain a curvature map;
performing gray stretching processing on the curvature map to obtain a gray stretched image;
carrying out image binarization processing on the gray level stretching image to obtain a binary image of the leather to be detected;
and processing the binary image of the leather to be detected by using a support vector machine, and outputting the detection result of the leather to be detected.
Further, as a preferred embodiment, the step of acquiring a plurality of first images captured under the irradiation of light sources with different positions includes the steps of:
fixing the position of the camera, and enabling the camera to be perpendicular to the leather detection platform;
arranging a plurality of light sources at different positions above the leather detection platform, and enabling an included angle between incident light of the light sources and the leather detection platform to be 30-60 degrees;
controlling a plurality of light sources to generate different illumination conditions;
controlling the camera to take a first image under a lighting condition;
and acquiring a first image obtained by shooting. Preferably, the number of first images is the same as the number of light sources, the number of light sources is 4, and/or the model number of the plurality of light sources is the same.
Further, as a preferred embodiment, the step of synthesizing the plurality of acquired first images into the second image by using the photometric stereo algorithm includes the following steps:
calculating a surface normal vector of the leather to be measured according to the plurality of collected first images;
calculating an orthogonal vector corresponding to the surface normal vector according to the calculated surface normal vector;
and generating a depth image according to the calculated orthogonal vector.
Further as a preferred embodiment, the step of obtaining the curvature map after performing curvature filtering on the second image specifically includes:
and performing curvature filtering on the second image by adopting a Gaussian curvature filter to obtain a curvature map.
Further, as a preferred embodiment, the expression formula of the gaussian curvature filter is as follows:
where U represents the second image of the input, H represents the curvature map of the output, and x represents the image convolution operation.
Further preferably, the step of performing gray-scale stretching processing on the curvature map includes the steps of:
obtaining the minimum gray value and the maximum gray value of the curvature map;
and stretching the gray value of the pixel point in the curvature map by using the minimum gray value and the maximum gray value.
Further as a preferred embodiment, in the step of performing gray stretching on the gray value of the pixel point in the curvature map by using the minimum gray value and the maximum gray value, a gray stretching formula adopted by the step is as follows:
wherein Y represents a gray value after gray stretching, X represents a gray value before gray stretching, and X representsminExpressed as the minimum gray value, XmaxExpressed as the maximum gray value.
Further preferably, the step of performing image binarization processing on the gray-scale stretched image uses a binarization conversion formula as follows:
threshold＝E(S(x))×3
where o (x) is expressed as an output binary image, s (x) is expressed as a pixel value of a pixel point in the grayscale stretched image, threshold is expressed as a threshold, and E (s (x)) is expressed as an expected value of the grayscale stretched image.
Example 2
The system corresponding to the above method, as shown in fig. 2, is a leather defect detection system based on photometric stereo vision, and the system comprises:
the first acquisition module is used for acquiring a plurality of first images shot under the irradiation action of light sources at different positions, wherein the first images refer to images of the leather to be detected;
the first processing module is used for synthesizing a plurality of acquired first images to obtain a second image by adopting a photometric stereo vision algorithm;
the second processing module is used for performing curvature filtering on the second image to obtain a curvature map;
the third processing module is used for performing gray stretching processing on the curvature map to obtain a gray stretching image;
the fourth processing module is used for carrying out image binarization processing on the gray level stretching image to obtain a binary image of the leather to be detected;
and the fifth processing module is used for processing the binary image of the leather to be detected by using the support vector machine and outputting the detection result of the leather to be detected. For the first acquisition module and the first to fifth processing modules, they may be program modules, or hardware circuits, or a combination of hardware and software.
Further as a preferred embodiment, the first acquisition module comprises:
the first control submodule is used for fixing the position of the camera and enabling the camera to be perpendicular to the leather detection platform;
the second control submodule is used for arranging the light sources at different positions above the leather detection platform and enabling an included angle between incident light of the light sources and the leather detection platform to be 30-60 degrees;
the third control sub-module is used for controlling the plurality of light sources to generate different illumination conditions;
a fourth control sub-module for controlling the camera to capture a first image under a lighting condition;
and the first acquisition submodule is used for acquiring a first image obtained by shooting.
Further as a preferred embodiment, the first processing module includes:
the first calculation submodule is used for calculating a surface normal vector of the leather to be measured according to the plurality of collected first images;
the second calculation submodule is used for calculating an orthogonal vector corresponding to the surface normal vector according to the calculated surface normal vector;
and the first generation submodule is used for generating a depth image according to the orthogonal vector obtained by calculation.
Further as a preferred embodiment, the second processing module is specifically configured to obtain a curvature map after performing curvature filtering on the second image by using a gaussian curvature filter. Preferably, the expression formula of the gaussian curvature filter described in this embodiment is the same as that of the gaussian curvature filter described in embodiment 1.
Further as a preferred embodiment, the third processing module includes:
the third calculation submodule is used for obtaining the minimum gray value and the maximum gray value of the curvature map;
and the first stretching submodule is used for carrying out gray stretching on the gray value of the pixel point in the curvature map by utilizing the minimum gray value and the maximum gray value.
Further as a preferred embodiment, the gray scale stretching formula used in the first stretching submodule is the same as the gray scale stretching formula used in example 1.
Further, as a preferred implementation, the binarization conversion formula adopted in the fourth processing module is the same as the binarization conversion formula adopted in embodiment 1.
Example 3
The device corresponding to the method is a leather defect detection device based on photometric stereo vision, and the device comprises:
a memory for storing at least one program;
a processor for loading the at least one program and executing the leather defect detection method steps based on photometric stereo as described in example 1 above.
Example 4
The invention will be further elucidated with reference to the preferred embodiment. In the present embodiment, the number of the first image/the third image and the number of the light sources are preferably 4.
The specific implementation steps of the leather defect detection method based on the photometric stereo vision are shown as follows.
The method comprises the following steps: the desired SVM is constructed. As shown in fig. 3, the step one specifically includes the following sub-steps.
S101, acquiring 4 third images shot under the irradiation action of light sources at different positions; the third image refers to an image of a leather defect detection sample;
for step S101, it includes the following steps:
s1011, fixing the position of the camera, and enabling the camera to be perpendicular to the leather detection platform;
s1012, arranging 4 light sources at different positions above the leather detection platform, and enabling an included angle between incident light of the light sources and the leather detection platform to be 30-60 degrees;
s1013, controlling the 4 light sources to generate different illumination conditions;
s1014, controlling the camera to shoot a third image under a lighting condition;
s1015, acquiring the shot third image, and acquiring 4 third images in total;
it can be seen that since the number of the third images is 4, 4 light sources should be controlled to generate 4 different lighting conditions; for the above-mentioned camera and 4 light sources, they form a photometric stereo experiment platform, as shown in fig. 4 specifically; in addition, as shown in fig. 4, the included angle between the incident light of the light source and the leather detection platform is preferably 40 °;
s102, synthesizing the 4 acquired third images by adopting a photometric stereo vision algorithm to obtain a fourth image with certain three-dimensional information;
for step S102, it includes the following steps:
s1021, calculating a surface normal vector of the leather defect detection sample according to the collected 4 third images;
firstly, the light intensity of any pixel point on the surface is expressed as: k isdN.L, wherein kdThe value is a constant, N is an unknown number (a surface normal vector to be solved), and L is a position parameter of the light source, so that pixel points of the 4 third images are respectively expressed as:
I1＝kd(Nx*L1x+Ny*L1y+Nz*L1y)
I2＝kd(Nx*L2x+Ny*L2y+Nz*L2y)
I3＝kd(Nx*L3x+Ny*L3y+Nz*L3y)
I4＝kd(Nx*LOx+Ny*L4y+Nz*L4y)
then let [ G ]]＝kd[N]Then the above equation set can be rewritten as N ═ G/kd；
Finally, using N ═ G/kdSolving to obtain a surface normal vector N;
s1022, calculating an orthogonal vector corresponding to the surface normal vector according to the surface normal vector calculated in the step S1021;
since N · V is 0 as known from the orthogonal relationship, the orthogonal vector V of the surface normal vector N can be obtained by solving the orthogonal vector V using the calculated surface normal vector N;
s1023, generating a depth image according to the orthogonal vector obtained by calculation; the depth image generated in step S1023 is the fourth image;
s103, curvature filtering is carried out on the fourth image by adopting a Gaussian curvature filter to obtain a first curvature map; wherein, the expression formula of the Gaussian curvature filter is as follows:
in the above formula, U represents the input image, H represents the output curvature map, and x represents the image convolution operation; the Gaussian curvature filter is adopted to carry out curvature filtering on the fourth image serving as the synthetic image, so that the calculation of curvature and complex geometric flow can be avoided, more importantly, the Gaussian curvature filter is very simple in structure and high in calculation speed, and in the method, the enhancement effect of a leather defect area can be improved on the basis of improving the processing efficiency, so that the accuracy of a subsequent leather defect detection result is higher;
s104, performing gray stretching processing on the first curvature map to obtain a first gray stretching image;
for step S104, it includes:
s1041, obtaining the minimum gray value X of the first curvature map by taking the gray value corresponding to each pixel point in the current first curvature map as XminAnd the maximum gray value Xmax；
Specifically, the gray value corresponding to each pixel point in the current first curvature map is X, the gray values of all the pixel points in the first curvature map are traversed, and the minimum gray value X of the first curvature map is obtainedminAnd the maximum gray value XmaxI.e. the corresponding minimum and maximum values in the image pixel values;
s1042, utilizing the minimum gray value XminAnd the maximum gray value XmaxPerforming gray stretching on the gray value of the pixel point in the first curvature map;
specifically, the gray value after the gray stretching is in the gray range of 0 to 255 is Y, and the calculation formula of Y is:
s105, performing image binarization processing on the first gray level stretching image to obtain a binary image of the leather defect detection sample, namely a second binary image; the binarization conversion formula adopted in the step is as follows:
threshold＝E(S(x))×3
wherein, o (x) is represented as an output binary image, s (x) is represented as a pixel value of a pixel point in a gray stretching image, threshold is represented as a threshold, and E (s (x)) is represented as an expected value of the gray stretching image, that is, an average gray value of the whole gray stretching image;
s106, training the SVM by adopting the second binary image and the image characteristics extracted from the second binary image; the image features extracted from the second binary image include, but are not limited to, features such as shapes, edges, area sizes, and the like;
for step S106, it includes:
s1061, taking the second binary image as training input data;
s1062, taking the image features extracted from the second binary image as training output data;
and S1063, training the SVM by using the training input data and the training output data, wherein the SVM obtained after training is the required SVM.
Step two: and carrying out defect detection processing on the leather to be detected by using the trained SVM so as to realize the positioning, detection and identification of the leather defects.
S201, collecting 4 first images shot under the irradiation action of light sources at different positions; the first image refers to an image of the leather to be detected;
for step S201, it includes the following steps:
s2011, fixing the position of the camera, and enabling the camera to be perpendicular to the leather detection platform;
s2012, arranging 4 light sources at different positions above the leather detection platform, and making an included angle between incident light of the light sources and the leather detection platform be 30-60 degrees, wherein the included angle is preferably 40 degrees;
s2013, controlling the 4 light sources to generate different illumination conditions;
s2014, controlling the camera to shoot a first image under a lighting condition;
s2015, collecting the shot first image to obtain 4 first images;
it can be seen that since the number of first images is 4, 4 light sources should be controlled to produce 4 different lighting conditions;
s202, synthesizing the 4 acquired first images by adopting a photometric stereo vision algorithm to obtain a second image with certain three-dimensional information;
for step S202, it includes the following steps:
s2021, calculating to obtain a surface normal vector of the leather to be measured according to the collected 4 first images;
s2022, calculating to obtain an orthogonal vector corresponding to the surface normal vector according to the surface normal vector calculated in the step S2021;
s2023, generating a depth image according to the orthogonal vector calculated in the step S2022; the depth image generated in step S2023 is the second image;
the implementation manner adopted by the above steps S2021 and S2022 is specifically as shown in the content of steps S1021 and S1022;
s203, curvature filtering is carried out on the second image by adopting the Gaussian curvature filter in the step S103 to obtain a second curvature map;
s204, performing gray stretching processing on the second curvature map to obtain a second gray stretching image;
for step S204, it includes:
s2041, obtaining the minimum gray value X of the second curvature map by taking the gray value corresponding to each pixel point in the current second curvature map as XminAnd the maximum gray value Xmax；
Specifically, the gray value corresponding to each pixel point in the current second curvature map is X, the gray values of all the pixel points in the second curvature map are traversed, and the minimum gray value X of the second curvature map is obtainedminAnd the maximum gray value XmaxI.e. the corresponding minimum and maximum values in the image pixel values;
s2042, utilizing the obtained minimum gray value XminAnd the maximum gray value XmaxPerforming gray stretching on the gray value of the pixel point in the second curvature map by using the calculation formula of Y in the step S1042, thereby obtaining a second gray stretched image;
s205, performing image binarization processing on the second gray scale stretching image to obtain a binary image of the leather to be detected, namely a first binary image; the binarization conversion formula adopted in this step is as shown in the formula in the above step S105;
s206, processing the binary image of the leather to be detected, namely the first binary image, by using the SVM trained in the step one, thereby outputting a defect detection result of the leather to be detected. As shown in fig. 5, the left four images are images (i.e. first images) of the leather to be detected input under different light source positions, the upper right side is a composite image (i.e. second image), and the lower right side is a schematic diagram of the detection result.
According to the invention, the photometric stereo vision experimental platform is used for collecting a plurality of images under the action of light sources at different positions and synthesizing a picture with certain three-dimensional information under the action of the photometric stereo vision technology, so that the synthesized picture has a very obvious enhancement effect and is beneficial to improving the precision of subsequent leather defect detection; and the curvature filtering has excellent filtering effect on the region with uneven surface, so that the curvature filtering is adopted on the basis of the synthetic image, the enhancement effect of the leather defect region can be improved on the basis of improving the processing efficiency, and the accuracy of the subsequent leather defect detection result is further improved.
The contents of the present embodiment are all applied to the above embodiments 1 to 3.
While the preferred embodiments of the present invention have been illustrated and described, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (8)
1. The leather defect detection method based on photometric stereo vision is characterized by comprising the following steps:
acquiring a plurality of first images shot under the irradiation action of light sources at different positions, wherein the first images refer to images of leather to be detected;
synthesizing a plurality of acquired first images by adopting a photometric stereo vision algorithm to obtain a second image;
curvature filtering is carried out on the second image to obtain a curvature map;
performing gray stretching processing on the curvature map to obtain a gray stretched image;
carrying out image binarization processing on the gray level stretching image to obtain a binary image of the leather to be detected;
processing the binary image of the leather to be detected by using a support vector machine, and outputting a detection result of the leather to be detected;
the step of acquiring a plurality of first images shot under the irradiation of light sources at different positions specifically comprises the following steps:
fixing the position of the camera, and enabling the camera to be perpendicular to the leather detection platform;
arranging a plurality of light sources at different positions above the leather detection platform, and enabling an included angle between incident light of the light sources and the leather detection platform to be 30-60 degrees;
controlling a plurality of light sources to generate different illumination conditions;
controlling the camera to take a first image under a lighting condition;
collecting a first image obtained by shooting
The step of synthesizing the plurality of acquired first images to obtain the second image by adopting the photometric stereo algorithm specifically comprises the following steps:
calculating a surface normal vector of the leather to be measured according to the plurality of collected first images;
calculating an orthogonal vector corresponding to the surface normal vector according to the calculated surface normal vector;
and generating a depth image according to the calculated orthogonal vector.
2. The leather defect detecting method based on photometric stereo as claimed in claim 1,
the step of obtaining the curvature map after curvature filtering is performed on the second image specifically includes:
and performing curvature filtering on the second image by adopting a Gaussian curvature filter to obtain a curvature map.
3. The leather defect detecting method based on photometric stereo as claimed in claim 2, wherein,
the expression formula of the gaussian curvature filter is as follows:
where U represents the second image of the input, H represents the curvature map of the output, and x represents the image convolution operation.
4. The leather defect detecting method based on photometric stereo as claimed in claim 1,
the step of performing gray scale stretching processing on the curvature map comprises the following steps:
obtaining the minimum gray value and the maximum gray value of the curvature map;
and stretching the gray value of the pixel point in the curvature map by using the minimum gray value and the maximum gray value.
5. The leather defect detecting method based on photometric stereo is characterized in that,
the step of stretching the gray value of the pixel point in the curvature map by using the minimum gray value and the maximum gray value is as follows:
wherein Y represents a gray value after gray stretching, X represents a gray value before gray stretching, and X representsminExpressed as the minimum gray value, XmaxExpressed as the maximum gray value.
6. The leather defect detecting method based on photometric stereo as claimed in claim 1,
the step of performing image binarization processing on the gray stretching image adopts a binarization conversion formula as follows:
threshold＝E(S(x))×3
where o (x) is expressed as an output binary image, s (x) is expressed as a pixel value of a pixel point in the grayscale stretched image, threshold is expressed as a threshold, and E (s (x)) is expressed as an expected value of the grayscale stretched image.
7. Leather defect detecting system based on luminosity stereovision, characterized in that, this system includes:
the first acquisition module is used for acquiring a plurality of first images shot under the irradiation action of light sources at different positions, wherein the first images refer to images of the leather to be detected;
the first processing module is used for synthesizing a plurality of acquired first images to obtain a second image by adopting a photometric stereo vision algorithm;
the second processing module is used for performing curvature filtering on the second image to obtain a curvature map;
the third processing module is used for performing gray stretching processing on the curvature map to obtain a gray stretching image;
the fourth processing module is used for carrying out image binarization processing on the gray level stretching image to obtain a binary image of the leather to be detected;
the fifth processing module is used for processing the binary image of the leather to be detected by using the support vector machine and outputting the detection result of the leather to be detected;
wherein the first acquisition module comprises:
the first control submodule is used for fixing the position of the camera and enabling the camera to be perpendicular to the leather detection platform;
the second control submodule is used for arranging the light sources at different positions above the leather detection platform and enabling an included angle between incident light of the light sources and the leather detection platform to be 30-60 degrees;
the third control sub-module is used for controlling the plurality of light sources to generate different illumination conditions;
a fourth control sub-module for controlling the camera to capture a first image under a lighting condition;
the first acquisition submodule is used for acquiring a first image obtained by shooting;
the first processing module comprises:
the first calculation submodule is used for calculating a surface normal vector of the leather to be measured according to the plurality of collected first images;
the second calculation submodule is used for calculating an orthogonal vector corresponding to the surface normal vector according to the calculated surface normal vector;
and the first generation submodule is used for generating a depth image according to the orthogonal vector obtained by calculation.
8. Leather defect detection device based on luminosity stereovision, its characterized in that, the device includes:
a memory for storing at least one program;
a processor for loading the at least one program and executing the leather defect detection method based on photometric stereo according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710847126.6A CN107607546B (en) | 2017-09-19 | 2017-09-19 | Leather defect detection method, system and device based on photometric stereo vision |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710847126.6A CN107607546B (en) | 2017-09-19 | 2017-09-19 | Leather defect detection method, system and device based on photometric stereo vision |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107607546A CN107607546A (en) | 2018-01-19 |
| CN107607546B true CN107607546B (en) | 2020-10-23 |
Family
ID=61061215
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710847126.6A Active CN107607546B (en) | 2017-09-19 | 2017-09-19 | Leather defect detection method, system and device based on photometric stereo vision |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107607546B (en) |
Families Citing this family (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108647554A (en) * | 2018-05-15 | 2018-10-12 | 佛山市南海区广工大数控装备协同创新研究院 | A kind of sport footwear air cushion stereoscopic two-dimensional code identification and detection device and method |
| CN110715939A (en) * | 2018-07-12 | 2020-01-21 | 卓峰智慧生态有限公司 | Artificial intelligence-based leather detection method and leather product production method |
| CN109523541A (en) * | 2018-11-23 | 2019-03-26 | 五邑大学 | A kind of metal surface fine defects detection method of view-based access control model |
| CN109829886A (en) * | 2018-12-25 | 2019-05-31 | 苏州江奥光电科技有限公司 | A kind of pcb board defect inspection method based on depth information |
| CN109932367A (en) * | 2019-03-14 | 2019-06-25 | 佛山缔乐视觉科技有限公司 | A kind of curved surface carved image acquisition device |
| CN109762949B (en) * | 2019-03-18 | 2021-04-27 | 香港纺织及成衣研发中心有限公司 | Fur processing method, device and equipment |
| CN110298835B (en) * | 2019-07-02 | 2023-07-18 | 广东工业大学 | Leather surface damage detection method, system and related device |
| CN110987954B (en) * | 2019-12-30 | 2021-10-22 | 江南大学 | Method and system for eliminating leather surface defect detection blind area |
| CN111624206B (en) * | 2020-07-03 | 2021-08-31 | 东北大学 | Metal surface defect detection method based on linear array camera stereoscopic vision |
| CN113155852B (en) * | 2021-04-08 | 2023-08-01 | 煤炭科学研究总院有限公司 | Detection method and device for transmission belt and electronic equipment |
| CN113092489B (en) * | 2021-05-20 | 2024-08-23 | 鲸朵(上海)智能科技有限公司 | System and method for detecting appearance defects of battery |
| CN113252567A (en) * | 2021-06-08 | 2021-08-13 | 菲特（天津）检测技术有限公司 | Method, system, medium and terminal for rapidly detecting multiple defects on surface of aluminum valve plate |
| CN113781424B (en) * | 2021-09-03 | 2024-02-27 | 苏州凌云光工业智能技术有限公司 | Surface defect detection method, device and equipment |
| CN116559181B (en) * | 2023-07-07 | 2023-10-10 | 杭州灵西机器人智能科技有限公司 | Defect detection method, system, device and medium based on luminosity stereoscopic vision |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102951151A (en) * | 2011-08-24 | 2013-03-06 | 现代摩比斯株式会社 | Lane maintaining auxiliary system for vehicles and method thereof |
| CN103592307A (en) * | 2012-08-17 | 2014-02-19 | 索尼公司 | 3d measuring device, 3d measuring method, program and method of manufacturing substrate |
| CN105787989A (en) * | 2016-03-18 | 2016-07-20 | 山东大学 | Measurement texture geometric feature reconstruction method based on photometric stereo |
-
2017
- 2017-09-19 CN CN201710847126.6A patent/CN107607546B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102951151A (en) * | 2011-08-24 | 2013-03-06 | 现代摩比斯株式会社 | Lane maintaining auxiliary system for vehicles and method thereof |
| CN103592307A (en) * | 2012-08-17 | 2014-02-19 | 索尼公司 | 3d measuring device, 3d measuring method, program and method of manufacturing substrate |
| CN105787989A (en) * | 2016-03-18 | 2016-07-20 | 山东大学 | Measurement texture geometric feature reconstruction method based on photometric stereo |
Non-Patent Citations (3)
| Title |
|---|
| 基于人脸图像的性别分类;陆庆庆;《中国优秀硕士学位论文全文数据库 信息科技辑》;20140715;15 * |
| 基于光度立体视觉的起皱织物表面形态重建研究;杨晓波等;《东华大学学报（自然科学版）》;20020430;第28卷(第2期);48-55 * |
| 杨晓波等.基于光度立体视觉的起皱织物表面形态重建研究.《东华大学学报（自然科学版）》.2002,第28卷(第2期),48-55. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107607546A (en) | 2018-01-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107607546B (en) | Leather defect detection method, system and device based on photometric stereo vision | |
| CN107845086B (en) | Method, system and device for detecting significant defects on leather surface | |
| CN112907530B (en) | Method and system for detecting disguised object based on grouped reverse attention | |
| CN105069423A (en) | Human body posture detection method and device | |
| CN108154519A (en) | Dividing method, device and the storage medium of eye fundus image medium vessels | |
| CN108734690A (en) | A kind of defects of vision detection device and its detection method | |
| CN112819772A (en) | High-precision rapid pattern detection and identification method | |
| CN110889837A (en) | Cloth flaw detection method with flaw classification function | |
| TW201516969A (en) | Visual object tracking method | |
| CN110738644A (en) | automobile coating surface defect detection method and system based on deep learning | |
| CN112288726B (en) | Method for detecting foreign matters on belt surface of underground belt conveyor | |
| CN107194432B (en) | Refrigerator door body identification method and system based on deep convolutional neural network | |
| JP2015141707A5 (en) | ||
| JP2004021373A (en) | Method and apparatus for estimating body and optical source information | |
| Fan et al. | Automatic IC character recognition system for IC test handler based on SVM | |
| CN112132135A (en) | Power grid transmission line detection method based on image processing and storage medium | |
| Nag et al. | Generating Vectors from Images using Multi-Stage Edge Detection for Robotic Artwork | |
| CN112464948A (en) | Natural scene target contour extraction method and system based on bionics | |
| CN113223099B (en) | RatSLAM environmental adaptability improving method and system based on biological vision model | |
| Pau et al. | Tiny defects identification of mechanical components in die-cast aluminum using artificial neural networks for micro-controllers | |
| CN117808808B (en) | Ore granularity detection method, system, electronic equipment and storage medium | |
| Bajrami et al. | From Dataset Creation to Defect Detection: A Proposed Procedure for a Custom CNN Approach for Polishing Applications on Low-Performance PCs. | |
| Khanal et al. | Fabric hairiness analysis for quality inspection of pile fabric products using computer vision technology | |
| EP2808843A3 (en) | Method of parameterisation of an image processing system for the monitoring of a machine tool | |
| CN107665487A (en) | Model organism IMAQ digitlization shape extraction reconstructing method |
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 | ||
| CB03 | Change of inventor or designer information | ||
| CB03 | Change of inventor or designer information |
Inventor after: Lin Jianfa Inventor after: Liu Gen Inventor after: Xiao Pan Inventor after: Huang Guancheng Inventor before: Lin Jianfa Inventor before: Liu Gen Inventor before: Wang Han Inventor before: Cai Nian Inventor before: Chen Xindu |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |