CN102175692A - System and method for detecting defects of fabric gray cloth quickly - Google Patents

Abstract

The invention relates to a system for detecting defects of fabric gray cloth quickly and automatically on line, which comprises a charge coupled device (CCD) camera, a computer with an image acquisition card and a stream line operation platform, wherein the stream line operation platform comprises a shielding device, two gray cloth clamp grooves and a cloth guiding rail; the CCD camera is fixed on the shielding device above the gray cloth, a lens is embedded into the shielding device, and a viewing field is aligned with the surface of the gray cloth; and a group of light-emitting diode (LED) light sources are fixed to the inner side of the bottom surface of the stream line operation platform and are arranged on the lower side of the gray cloth. In the system, the acquisition, processing and characteristic extraction of fabric gray cloth images and the judgment and control of image data are performed by the computer, so that a final result is sent to an alarm engine bell by output equipment, and the classification and statistics of the defects of the fabric gray cloth are performed. Therefore, the automated, objective, non-contact, high-accuracy and quick detection of the defects of the fabric gray cloth is realized.

Description

Fabric grey cloth fault rapid detection system and method

Technical field

The present invention relates to a kind of detection system, relate in particular to a kind of on-line automatic rapid detection system of fabric grey cloth fault and method, the special employing obtained the grey cloth surface image and carried out Digital Image Processing and detect grey quality, and fault discerned and classifies, testing result accurately, efficient, automaticity is high.

Background technology

In the textile production, quality control and detection are very important, and it is most important part wherein that fabric defects detects.Up to now, traditional fabric defects detection is finished by artificial vision's offline inspection.But because artificial perching is a kind of dullness, dull and onerous toil, percher multipotency in a hour is found 200 faults, artificial perching concentrated force at most also can only be kept 20～30 minutes, surpass this time workman and can produce fatigue, perching speed only is 5～20m/min, surpass this speed and omission and flase drop will occur, and, can't solve continuously and stably by human eye for the accurately quick measurement of this type of fine size, the quick coupling of pattern and the vision-based detection problems such as quick identification of target at all.Therefore Automatic Detection of Fabric Defects becomes one of common heat subject of paying close attention to and studying of Chinese scholars in recent years.Become the crossing research field in the forward position that the scholar of weaving subject and information subject plays an active part in based on the Automatic Detection of Fabric Defects of computer vision technique, though obtained certain achievement, wherein most of or be in the news with forms such as algorithm or experimental prototypes.

Because machine vision is emphasized practicality, requirement can adapt to the industry spot rugged environment, more emphasizes real-time, requires high-speed and high precision, so be specially adapted to the quality testing in the high volume production process.Utilize computer vision that fabric defects is detected automatically, not only have characteristics simple in structure, low-cost, that detection speed is fast, precision is high, and for the kind of fabric defects, size and objective standard is formulated in the assessment of product quality laid the first stone.

In recent years, the domestic upsurge that development and exploitation Automatic Detection of Fabric Defects system have also occurred, but present research still is in the starting stage.The technology of present stage mainly is based on the machine vision method of Digital Image Processing and carries out fabric grey cloth defect detection, detects by the extraction of eigenwert and the discriminator algorithm of fault.At present, also there are some problem demanding prompt solutions in the weaving online measuring technique based on machine vision: 1) design theory and method are single, can't satisfy in real time and online detection requirements at a high speed, and also be in the laboratory development state mostly; 2) high speed processing of image reaches bottleneck, and collection and the processing at line image occupied a large amount of time at a high speed, influences the online detection of cloth; 3) lacking the high-speed image Processing Algorithm, is improvement and the processing to original algorithm mostly.The transplantability and the robustness of algorithm are poor.

Summary of the invention

The object of the present invention is to provide a kind of automatically, accurately and rapidly based on the fabric grey cloth on-line automatic monitoring system and the method for Digital Image Processing.The present invention utilizes the feature of grey cloth fault, adopt ccd video camera to obtain the grey cloth surface image and handle, judge, thereby realize online, in real time the fabric grey quality is detected, and realize the classification of grey cloth fault.

In order to achieve the above object, the present invention adopts following technical proposals:

The online fast automatic detecting of a kind of fabric grey cloth system comprises a ccd video camera, a computing machine and a stream line operation platform that is equipped with image pick-up card.The stream line operation platform comprises a shield assembly, and two grey cloth draw-in grooves are located at the selvage guide rail of shield assembly both sides and guiding grey cloth respectively.Ccd video camera is fixed in the shield assembly end face, and camera lens embeds in the shield assembly, and the grey cloth surface is aimed in the visual field.One group of led light source is fixed in inboard, stream line operation platform bottom surface, promptly is positioned at fabric grey cloth below.Grey cloth can pass from two draw-in grooves of shield assembly both sides.Computing machine is connected with ccd video camera by image pick-up card and is connected with the warning alarm bell by outlet terminal.The fault handling procedure of one cover according to the compiling of grey cloth surface characteristics is housed in the computing machine; With different fabric grey cloth surface abstract be a group system parameter, obtain the grey cloth surface image by ccd video camera, utilize computing machine the grey cloth surface image to be carried out analyzing and processing by program compiler.

The effect of system's critical piece:

Ccd video camera: online acquisition fabric grey cloth image, and be transferred to computing machine through connecting line.

Led light source: because one group of led light source can provide evenly the illumination of low reflection.Light source is provided on the one hand the sealing photographed scene, may prevents that on the other hand uneven illumination from influencing picture quality.

The selvage guide rail: guiding fabric grey cloth is with certain speed, and assigned direction at the uniform velocity advances.

Grey cloth draw-in groove: prevent because the grey cloth material flexibility is strong, and length is big, and the edge curls inward causes the amplitude deformation on grey cloth surface, thereby influence grabs the effect on grey cloth surface, increases the difficulty and the error that detect.It is positioned at the shield assembly both sides.

Shield assembly: because ccd video camera is for illumination sensitivity very, shielding can prevent the illuminated the way instability of image pixel of ambient light, increases the precision of detection.

Computing machine: mainly the grey cloth image that CCD is collected is analyzed, and obtains fault information, carries out fault classification and statistics.

The online fast automatic detecting method of a kind of fabric defects, make grey cloth under the drive of selvage guide rail, at the uniform velocity pass through the draw-in groove of shield assembly both sides, entering in the ccd video camera visual field slowly then, through the computing machine of assembling image pick-up card to image gather, processing and feature extraction, and view data judged and control, final result is sent to warning by output device hold up bell, and carry out the classification and the statistics of grey cloth fault.Thereby realize robotization, objective, contactless, high precision, online detection fast.

Computing machine is as follows to grey cloth surface image handling procedure step:

1. from image pick-up card, read in a frame image data;

2. image is carried out color conversion;

3. image filtering;

4. Threshold Segmentation;

5. morphology is handled;

6. feature extraction;

7. fault classification and output result or warning.

Repeating step 1-7 realizes that the online streamline of fabric grey cloth fault detects.

System and method of the present invention is compared with existing fabric defects surface inspecting method, is the online fast automatic detecting technology of a kind of fabric defects based on monocular vision, has following conspicuous outstanding substantive distinguishing features and remarkable advantage:

1. use ccd video camera as sensor, obtain the grey cloth surface, really realized automatic detection thereby replace naked eyes.

2. because CCD has certain distance apart from grey cloth, thus can realize non-contact detection, thus prevent that swiping grey cloth surface from causing the grey cloth surface distress.

3. use various image processing algorithms, the single-frame images in the image sequence is carried out grey cloth detect.

4. has stronger adaptability.Abstract to different faults is one group of fault parameter, can adapt to different fabric grey cloths by simple modification environmental parameter and detect, and makes the range of application of system become wide.

5. accuracy of detection height, speed are fast, real-time.Native system utilizes ccd video camera to obtain image, and by Computer Analysis, processing and judgement, all operations all is an automation process, thereby can realize in real time, fast, accurately detecting fault.

Description of drawings

Description of drawings

Fig. 1 is the one-piece construction synoptic diagram of system of the present invention.

Fig. 2 is the process flow diagram that fabric grey cloth surface image of the present invention is analyzed.

Embodiment

Describe technical scheme of the present invention in detail below in conjunction with accompanying drawing.

As shown in Figure 1, the structural representation of the online fast automatic detecting of fabric defects of the present invention system, comprise a ccd video camera (1), a computing machine (8) and a stream line operation platform (7) that is equipped with image pick-up card, described stream line operation platform is equipped with the selvage guide rail (3) of guiding grey cloth, two grey cloth draw-in grooves (6) and a shield assembly (4), described ccd video camera (1) is fixed on the shield assembly (4) directly over the fabric grey cloth (5), vision alignment grey cloth (5) surface, the camera lens of ccd video camera (1) embeds in the shield assembly (4), and its field range just in time is the drop shadow spread of shield assembly on surface level.One group of led light source (2) is fixed in the below of fabric grey cloth (5), and fabric grey cloth (5) can pass from two draw-in grooves (6) of the shield assembly both sides on the operating platform; Described computing machine (8) is connected with described ccd video camera (1) and is connected with warning alarm bell (9) by outlet terminal by its image pick-up card; The fault handling procedure of one cover according to the compiling of grey cloth surface characteristics is housed in the described computing machine (8); With different fabric grey cloth (5) surface abstract be a group system parameter, obtain the grey cloth surface image by ccd video camera (1), utilize computing machine (8) the grey cloth surface image to be carried out analyzing and processing by program compiler, obtain final testing result.

Fig. 2 is the process flow diagram that the fabric grey cloth online automatic detection method based on image of present embodiment adopts said system to detect.Mainly depend on image processing method based on the detection system of machine vision and realize that the implementation method of system of the present invention also is based on image processing techniques and realizes testing goal.Mainly comprise image filtering, cut apart, content such as feature extraction, image recognition and understanding.After these processing, the quality of output image improves, and is convenient to computing machine image is further analyzed, handled and discern, judges.Based on image processing method, in the system computer of the present invention be to grey cloth surface image handling procedure step (as Fig. 2):

1.CCD video camera is to gather in the mode of frame/field, requirement (within 1/4 millimeter of the error, speed per second 5-6 rice) and ccd video camera hardware (resolution 768 * 576 in view of the product of production line quality, frame per second: the length that basis 10 frame/seconds), ccd video camera single frames obtain grey cloth is 0.6 meter.

2. color conversion.The image of ccd video camera collection is a coloured image, need convert gray level to and handle, and utilizes formula (1) to change.

nGray＝0.299×R+0.587×G+0.114×B (1)

Wherein, R, G, B are respectively the red, green, blue component in the coloured image, and nGray is the gray-scale value after changing.

3. image filtering.Fabric grey cloth image may be owing to optical system causes distortion, noise pollution etc. when gathering, and makes image blurringly, so must these images that degrades be improved, can reach the purpose of improving image by filtering.What native system adopted is the standard medium filtering.It is a kind of nonlinear properties disposal route, is preconditioning technique the most frequently used in the image processing techniques.If the set of the pixel grey scale of two dimensional image is { X _{I, j}(i, j) ∈ Z ², Z ²It is two-dimentional set of integers.For size is that pixel value intermediate value in the window of A=m * n (containing odd number of pixels) is defined as

$Y_{i,j}=\underset{A}{\mathrm{Median}}[X_{i+k,j+l}(k,l)\∈A]---\left(2\right)$

Following formula is represented the odd number of pixels in the window to get the intermediate pixel value and to compose to Y by the big minispread of gray-scale value _{I, j}, then with Y _{I, j}Replace center pixel value among the two-dimentional window A as the output of medium filtering.On image this window by from left to right, order mobile window from top to bottom.

4. image segmentation.Image segmentation will not have the image section of fault and will reject only remaining fault part as a setting.Native system adopts big Tianjin method to realize.Big Tianjin method is called the Ostu method again, is a kind of automatic threshold dividing method that makes the inter-class variance maximum.This method has simply, the fast characteristics of processing speed, is a kind of selection of threshold method commonly used.It is derived on discriminatory analysis principle of least square method basis and draws.Its basic thought is as follows: establishing the image pixel number is N, and the gray-scale value scope is [0, L-1], and the corresponding grey scale level is that the number of pixels of k is n _k, probability is:

p(k)＝n _k/N k＝0，1，2，Λ，L-1

$\underset{k=0}{\overset{L-1}{\mathrm{\Σ}}}p\left(k\right)=1---\left(3\right)$

Pixel in the image is divided into two class C by gray-scale value with threshold value T ₀And C ₁, C ₀Form C by the pixel of gray-scale value between [0, T] ₁Form by the pixel of gray-scale value between [T+1, L-1], for intensity profile probability, C ₀And C ₁The probability and the average that occur are respectively:

${\mathrm{\ω}}_0=\underset{k=0}{\overset{T}{\mathrm{\Σ}}}p\left(k\right),$ ${\mathrm{\ω}}_1=\underset{k=T+1}{\overset{L-1}{\mathrm{\Σ}}}p\left(k\right)=1-{\mathrm{\ω}}_0---\left(4\right)$

${\mathrm{\μ}}_0=\underset{k=0}{\overset{T}{\mathrm{\Σ}}}\mathrm{kp}\left(k\right)/{\mathrm{\ω}}_0,$ ${\mathrm{\μ}}_1=\underset{k=T+1}{\overset{L-1}{\mathrm{\Σ}}}\mathrm{kp}\left(k\right)/{\mathrm{\ω}}_1---\left(5\right)$

The average that can be got entire image by top formula (4) and (5) is:

μ＝ω ₀μ ₀+ω ₁μ ₁ (6)

C ₀And C ₁The variance of class can be obtained by following formula:

${\mathrm{\σ}}_0^2=\underset{k=0}{\overset{T}{\mathrm{\Σ}}}{(k-{\mathrm{\μ}}_0)}^{2}p\left(k\right)/{\mathrm{\ω}}_0$ (7)

${\mathrm{\σ}}_1^2=\underset{k=T+1}{\overset{L-1}{\mathrm{\Σ}}}{(k-{\mathrm{\μ}}_1)}^{2}p\left(k\right)/{\mathrm{\ω}}_1$

Inter-class variance is defined as:

${\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="B"\; filenames="HSA00000453396000011.png"\; state="\{\{state\}\}">B</figure-callout>}}^2={\mathrm{\&omega;}}_0{({\mathrm{\&mu;}}_0-\mathrm{\&mu;})}^2+{\mathrm{\&omega;}}_1{({\mathrm{\&mu;}}_1-\mathrm{\&mu;})}^2$ (8)

$={\mathrm{\&omega;}}_0{\mathrm{\&omega;}}_1{({\mathrm{\&mu;}}_0-{\mathrm{\&mu;}}_1)}^2$

Threshold value T is divided into C with image ₀And C ₁Two parts allow T in [0, L-1] scope value successively, make

Maximum T value is the optimal threshold of Ostu method.

5. morphology is handled.Through step 4,, can see fault shape size significantly if the image of gathering has fault.But, having tangible isolated point in the image of cutting apart, it does not belong to the fault part, but owing to noise causes, can adopt morphologic corrosion and expand and eliminate this part isolated point.

With structural elements S the erosion operation that bianry image B carries out is expressed as

Be defined as follows:

$\mathrm{B\&Theta;S}=\left\{b\right|b+s\&Element;B\&ForAll;s\&Element;S\}---\left(9\right)$

With structural elements S the dilation operation that bianry image B carries out is expressed as

Be defined as follows:

$B\&CirclePlus;S=\underset{b\&Element;B}{Y}{S}_b---\left(10\right)$

Native system adopts the method for corroding after expansion earlier to handle, and used structural unit is 3 * 3, can obtain apparent fault image at last.

6. feature extraction

Calculate the geometric properties of the fault image that obtains through step 5.Comprise: area, shape, horizontal diameter and perpendicular diameter ratio and reciprocal.Simultaneously, can in former gray level image, locate defect position and size in the relevant position.To non-defect position, can extract the grey cloth image of one 128 * 128 pixel or 64 * 64 pixel sizes, calculate its gray average.With the difference of the gray average of the gray average of non-fault and website part a feature as the grey cloth fault.

7. fault classification

According to the fault eigenwert of extracting, carry out the fault classification.

The feature of broken hole.Shape: sub-elliptical; Ratio＜7 of horizontal diameter/perpendicular diameter; Poor＜50 of gray-scale value.

The feature of greasy dirt.Shape: approximate circle or ellipse; Ratio＜7 of horizontal diameter/perpendicular diameter; Poor＞100 of gray-scale value.

The feature of tieing.Shape: approximate circle; Ratio＜3 of horizontal diameter/perpendicular diameter; Poor＜20 of gray-scale value.

The feature of crapand.Shape: approximate rectangular; Ratio＞10 of horizontal diameter/perpendicular diameter; Poor＜50 of gray-scale value.

System obtains continuous single-frame images by ccd video camera, through the single frames collection, obtains piece image.By the image software analysis tool, this width of cloth image is handled then, and is realized fault identification and statistic of classification, thus realize to fabric grey cloth surface in real time, fast, efficiently, robotization accurately detects.

Claims (3)

1. the online fast automatic detecting of fabric grey cloth system, its principal character is to comprise

A ccd video camera;

A stream line operation platform comprises a shield assembly, and two grey cloth draw-in grooves are located at the selvage guide rail of shield assembly both sides and guiding grey cloth respectively;

A computing machine is equipped with image pick-up card;

One group of led light source is fixed in inboard, described stream line operation platform bottom surface, places below the grey cloth;

Described ccd video camera is fixed in the shield assembly end face, and camera lens embeds in the shield assembly, and the grey cloth surface is aimed in the visual field;

Described computing machine is connected with described ccd video camera by described image pick-up card and is connected with the warning alarm bell by outlet terminal.

2. online fast automatic detecting method of fabric grey cloth, it is characterized in that grey cloth under the drive of selvage guide rail at the uniform velocity by being located at the draw-in groove of shield assembly both sides, described grey cloth entering in the ccd video camera visual field slowly, through the computing machine of assembling image pick-up card to image gather, processing and feature extraction, and view data judged and control, final result is sent to warning by output device hold up bell, and carry out the classification and the statistics of grey cloth fault.

3. the online fast automatic detecting method of fabric grey cloth according to claim 2 is characterized in that described computing machine is as follows to grey cloth image processing program step:

1. from image pick-up card, read in a frame image data;

2. image is carried out color conversion;

3. image filtering;

4. Threshold Segmentation;

5. morphology is handled;

6. feature extraction;

7. fault classification and output result or warning.

Images