A kind of flaw online test method suitable for Woven textiles
Technical field
The present invention relates to a kind of flaw online test method suitable for Woven textiles, belong to textile flaw and know online
Other technical field.
Background technology
In Modern Textile Industry, can be carried instead of the automatic detection that human eye carries out fabric defects with advanced detection technique
High detection efficiency, the quality for reducing labour, reducing labour intensity and further improve fabric.In China, the big portion of Fabric Detection
Divide or completed by artificial vision, in detection process, human eye vision has deviation, and fidelity factor is than relatively low and testing result not
Stabilization, it is incompatible with large-scale industrial production.In textile production, if flaw can not be found in time, will produce
Some substandard products cloth, cause the waste of material and the energy.If installing fabric defect visual inspection system, it is found that flaw can be carried out
Treatment, to reduce the waste of material.
Retrieved through applicant and found, Chinese invention patent application CN102967606A, it is proposed that one kind weaving woven fabric flaw
Defect vision detection system, its operation principle is that the feature extraction of cloth photo is carried out using image recognition technology, after comparison, will
Result is sent to host processing systems, completes to examine whole machine vision the control of distribution system through host processing systems, and
And corresponding final process is made to the result that it feeds back in real time.The technology can real-time detection go out fabric defects point, and send
Alarm number, stops weaving loom operation, and guiding workman processes current flaw, be also effectively reduced labor strength and labour into
This.
The detection that image recognition technology is applied to fabric is had become the customary means of this area, but existing identification side
Fado is the direct conversion of universal method, and its specific aim is not strong, however it remains missing inspection and false retrieval, it is impossible to meet wanting for high-quality
Ask.The present invention is identified emphatically according to the texture feature of fabric to the flaw of fabric.
The content of the invention
The purpose of the present invention:Overcome the defect of above-mentioned prior art, propose a kind of new flaw suitable for Woven textiles
The characteristics of defect online test method is for fabric carries out flaw identification, makes missing inspection, fallout ratio reduction.
Flaw online test method suitable for Woven textiles proposed by the present invention, it is characterised in that:Cloth advance side
What is be spaced upward is provided with upstream camera, downstream camera, is set according to cloth pace and upstream and downstream camera spacing
The shooting time for putting two cameras is poor, it is ensured that two cameras can photograph the same region of Woven textiles, be connected
Continuous is some to image, and this method includes following steps:
1st step:Using upstream, downstream camera captured in real-time Woven textiles, and paired image is transferred to calculate
Machine;
2nd step:Computer is pre-processed to the image for receiving, including:Gray level image is normalized, and is passed through
Gray scale stretching strengthens picture contrast;
3rd step:The image zooming-out textural characteristics vector shot to upstream camera, comprises the following steps that:
A1, dual-tree complex wavelet transform is carried out to image, obtain 6 256 multiply 256 matrix;
A2,6 matrixes that will be obtained substitute into equine husband's model, ask for the parameter of equine husband's model, and the parameter to obtaining
It is normalized;
A3, to after normalization parameter build vector, obtain the textural characteristics vector of fabric in the image;
4th step:The textural characteristics vector that 3rd step is obtained is compared with the textural characteristics vector of qualified products, if
Both Euclidean distances are not more than default qualified products threshold value, then the fabric in the image is qualified products, goes to the 1st step;
The image that otherwise downstream camera is shot obtains two pieces of subgraphs along center wire cutting;
5th step:Two pieces of textural characteristics vectors of subgraph are extracted respectively using the method for the 3rd step, and respectively by two pieces
The textural characteristics vector of subgraph is compared with the textural characteristics vector of qualified products, if both Euclidean distances are little
In default qualified products threshold value, then the fabric in the image is qualified products, then go to the 1st step;If both Euclidean away from
From more than qualified products threshold value, then by corresponding subgraph textural characteristics vector respectively with the woven fabric lines of all pre-selection flaw species
Reason characteristic vector is compared, woven in corresponding subgraph if both Euclidean distances are not more than corresponding flaw threshold value
Thing belongs to corresponding flaw species, and computer is recorded;If both Euclidean distances are all higher than corresponding flaw threshold value, phase
Answer the woven fabric in subgraph for other flaw species, computer is recorded, and halt instruction is sent according to above-mentioned flaw species,
And point out corresponding personnel to be processed immediately.
Further improvement of the invention is as follows:
1st, the above-mentioned flaw online test method suitable for Woven textiles, the pre-selection flaw species bag in the 5th step
Include:Lack warp, crapand, again stain, warp, weight latitude, broken hole.
2nd, the textural characteristics vector and qualified products threshold value determination method of the qualified products are as follows:
B1, the Woven textiles sample of at least 200 qualified products of selection carry out IMAQ;
B2, the textural characteristics vector that each image is extracted using the method in the 3rd step, all qualified products image textures
The cluster centre of characteristic vector is qualified products textural characteristics vector center vector;
B3, the textural characteristics vector of all qualified products images are to the qualified products textural characteristics vector center vector
The maximum of Euclidean distance is qualified products threshold value.
3rd, the woven fabric textural characteristics vector of flaw and corresponding flaw Threshold are as follows:
C1, the Woven textiles sample of at least 200 specified flaw categories of selection carry out IMAQ;
C2, the textural characteristics vector that each image is extracted using the method in the 3rd step, all specified flaw categories
The cluster centre of image texture characteristic vector is the center vector of corresponding flaw product textural characteristics vector;
C3, the textural characteristics vector of all such flaw product images are to the center of corresponding flaw product textural characteristics vector
The maximum of the Euclidean distance of vector is corresponding flaw threshold value.
4th, the shooting time difference T=S/V of two cameras, S is that camera shoots the distance between center in formula, and V is machine
The pace of textile is knitted, downstream camera shooting time is later than upstream camera shooting time.
The present invention is carried using dual-tree complex wavelet for the texture feature of Woven textiles with the feature of equine husband's models coupling
The method of taking carries out texture blending, compared to traditional texture blending method, can quickly, effectively, the accurate Woven textiles that extract
Textural characteristics, and the textural characteristics for extracting are more beneficial for identification, the results showed significantly improving the accuracy of identification.This hair
Bright to carry out fabric collection using dual camera, the monoblock cloth for taking the photograph region to upstream camera first carries out texture blending, just
Step detects whether to be qualified products, if unqualified take the photograph the secondary texture blending of cloth subregion by downstream camera again, and carries out
Flaw is recognized.The present invention is dexterously shot in diverse location using two cameras to same cloth, it is to avoid due to anti-
Erroneous judgement caused by the interference such as light, subregional secondary identification is carried out after preliminary identification, substantially increases identification accuracy, while
Detection speed is can guarantee that again, and occupancy resource is smaller, can adapt to on-line checking, does not shut down detection.
To sum up, the present invention greatly reduces identification error rate, improves the accuracy of flaw database, contributes to the producer couple
Weaving parameter is adjusted to improve product quality.
Brief description of the drawings
The present invention is further illustrated below in conjunction with the accompanying drawings.
Fig. 1 is the flow diagram of the inventive method.
Specific embodiment
The present invention will be further described with specific embodiment below in conjunction with the accompanying drawings.
The present embodiment is applied to the flaw online test method of Woven textiles, and its improvement is:In cloth direction of advance
What side was spaced is provided with upstream camera, downstream camera, and two are set according to cloth pace and upstream and downstream camera spacing
The shooting time difference T of camera, it is ensured that two cameras can photograph the same region of Woven textiles, and shooting time is poor
T=S/V, S is that camera shoots the distance between center in formula, and V is the pace of Woven textiles, downstream camera
Shooting time is later than upstream camera shooting time, obtains continuous some to image;The present embodiment method flow diagram is shown in Fig. 1,
Comprise the following steps that:
1st step:Using upstream, downstream camera captured in real-time Woven textiles, and paired image is transferred to calculate
Machine;
2nd step:Computer is pre-processed to the image for receiving, including:Gray level image is normalized, and is passed through
Gray scale stretching strengthens picture contrast;
3rd step:The image zooming-out textural characteristics vector shot to upstream camera, comprises the following steps that:
A1, dual-tree complex wavelet transform is carried out to image, obtain 6 256 multiply 256 matrix;
A2,6 matrixes that will be obtained substitute into equine husband's model, ask for the parameter of equine husband's model, and the parameter to obtaining
It is normalized;
A3, to after normalization parameter build vector, obtain the textural characteristics vector of fabric in the image;
4th step:The textural characteristics vector that 3rd step is obtained is compared with the textural characteristics vector of qualified products, if
Both Euclidean distances are not more than default qualified products threshold value, then the fabric in the image is qualified products, goes to the 1st step;
The image that otherwise downstream camera is shot obtains two pieces of subgraphs along center wire cutting;
5th step:Two pieces of textural characteristics vectors of subgraph are extracted respectively using the method for the 3rd step, and respectively by two pieces
The textural characteristics vector of subgraph is compared with the textural characteristics vector of qualified products.If both Euclidean distances are little
In default qualified products threshold value, then the fabric in the image is qualified products, then go to the 1st step.If both Euclidean away from
From more than qualified products threshold value, then by corresponding subgraph textural characteristics vector respectively with the woven fabric lines of all pre-selection flaw species
Reason characteristic vector is compared, woven in corresponding subgraph if both Euclidean distances are not more than corresponding flaw threshold value
Thing belongs to corresponding flaw species, and computer is recorded;If both Euclidean distances are all higher than corresponding flaw threshold value, phase
Answer the woven fabric in subgraph for other flaw species, computer is recorded to flaw species and sent halt instruction, and is carried
Show that corresponding personnel are processed immediately.
Pre-selection flaw species in 5th step includes:Lack warp, crapand, again stain, warp, weight latitude, broken hole.
The textural characteristics vector and qualified products threshold value determination method of this implementation qualified products are as follows:
B1, the Woven textiles sample of at least 200 qualified products of selection carry out IMAQ;
B2, the textural characteristics vector that each image is extracted using the method in the 3rd step, all qualified products image textures
The cluster centre of characteristic vector is qualified products textural characteristics vector center vector;
B3, the textural characteristics vector of all qualified products images are to the qualified products textural characteristics vector center vector
The maximum of Euclidean distance is qualified products threshold value.
The woven fabric textural characteristics vector of the present embodiment flaw and corresponding flaw Threshold are as follows:
C1, the Woven textiles sample of at least 200 specified flaw categories of selection carry out IMAQ;
C2, the textural characteristics vector that each image is extracted using the method in the 3rd step, all specified flaw categories
The cluster centre of image texture characteristic vector is the center vector of corresponding flaw product textural characteristics vector;
C3, the textural characteristics vector of all such flaw product images are to the center of corresponding flaw product textural characteristics vector
The maximum of the Euclidean distance of vector is corresponding flaw threshold value.
In addition to the implementation, the present invention can also have other embodiment.All use equivalents or equivalent transformation shape
Into technical scheme, all fall within the protection domain of application claims.