CN106097312A - A kind of glove based on machine vision tear and greasy dirt detection method - Google Patents

Abstract

The present invention relates to a kind of glove based on machine vision tear and greasy dirt detection method, for detecting the glove on production line, comprise the following steps: (1) uses industrial camera to shoot several continuous pictures of multiple glove；(2) utilize Kalman filter opponent set to carry out visual tracking, obtain the image information of multiple angles of same glove；(3) opponent packs into row contours extract, and is cut into suitable size；(4) whether detection glove tear；(5) whether detection glove have greasy dirt.Compared with prior art, the present invention has the advantages such as easy to use.

Description

A kind of glove based on machine vision tear and greasy dirt detection method

Technical field

The present invention relates to a kind of glove defect inspection method, especially relate to a kind of glove based on machine vision tear and Greasy dirt detection method.

Background technology

In current manufacturing business, the defects detection for product is completed by manual operation mostly.So Do and resulted in the problem that owing to people can be tired, so cannot ensure that the quality of detection has been at higher level；It addition, work The environment that industry produces is relatively more severe, and human body exists certain harm.During glove produce, due to raw materials for production and adding The reason of work technique, the detection workman of production scene is exposed to high temperature, among noisy environment, takes the little hour wheel of each two once The mode on hilllock carries out the defects detection of product.Either for workman, or for enterprise, a kind of effective manner is taked to replace It is imperative to carry out detecting for human eye.

Defects detection based on machine vision has had and has been widely applied very much, is the technology of a relative maturity, in reality There have been some successful stories on border in producing.But, current defects detection object is nearly all to have fixing geometric form The rigid objects of shape, from the defects detection for steel surface started most, lacking for glass, even fruit finally Sunken detection broadly falls into such category.And flexible article such for glove, owing to the change of shape is very big, cause glove Difference between individuality is the biggest, brings certain difficulty and challenge for defects detection.

Summary of the invention

Defect that the purpose of the present invention is contemplated to overcome above-mentioned prior art to exist and provide a kind of efficiency high based on The glove of machine vision tear and greasy dirt detection method.

The purpose of the present invention can be achieved through the following technical solutions: a kind of glove based on machine vision tear and oil Dirty detection method, for detecting the glove on production line, comprises the following steps:

(1) industrial camera is used to shoot several continuous pictures of multiple glove；

(2) utilize Kalman filter opponent set to carry out visual tracking, obtain the image of multiple angles of same glove Information；

(3) opponent packs into row contours extract, and is cut into suitable size；

(4) whether detection glove tear；

(5) whether detection glove have greasy dirt.

In described step (1), during shooting, increase black background on a production line.

Described step (2), particularly as follows: the entirety of glove and fingerprint is regarded as a rectangle, takes this rectangular centre as card The input of Thalmann filter, estimates the position of these glove in next frame image according to this input, obtains the multiple of same glove The image information of angle, it is achieved tracing task.

Described step (3), particularly as follows: use Canny edge detection operator, finds the highest, minimum, of glove profile Four left and the rightest point one rectangles of structure, the scope cutting covered by this rectangle is out, to be processed as next step Picture.

Described step (4) particularly as follows:

(401) whether color, if it has, then glove tear, otherwise go to step if judging to have powerful connections in the picture of glove profile (402)；

(402) judge in addition to glove stub area, if having the color of glove layer material superposition background colour, if it has, Then glove tear, otherwise, glove without tearing, described glove stub area with glove edge as partial contour, and only one The color of layer glove material superposition background colour.

Described step (5) particularly as follows:

(501) according to formula, the color space of picture is transformed into HSV from RGB,

$s=\left\{\begin{array}{cc}0,& \begin{array}{cc}if& max=0\end{array}\\ \frac{max-min}{max}=1-\frac{min}{\max },& otherwise\end{array}\right.$

V=max

Wherein, r, g, b are three parameter transform in artwork rgb color space to the value in [0,1] interval, max respectively Being r, the maximum in g, b, min is r, the minima in g, b, and h represents that tone, s represent that saturation, v represent lightness；

(502) if 40≤h≤50, then having greasy dirt on glove, otherwise, glove are without greasy dirt.

Before opponent packs into row contours extract in described step (3), erosion algorithm and expansion algorithm is used to eliminate picture In noise.

Compared with prior art, the invention have the advantages that

(1) Kalman filter opponent set is used to be tracked, because Kalman filter can be by when former frame figure The position of glove during the location estimation of glove goes out next frame image in Xiang, compares the change of glove location in direct comparison two frame picture Change the method being tracked, it is possible to obtain follow the tracks of result more accurately and have more preferable robustness for noise；

(2) using Canny edge detection operator, opponent packs into row contours extract, in the feelings ensureing that photo resolution is certain Under condition, the size of picture is minimum, improves the speed of defects detection；

(3), before opponent packs into row contours extract, use erosion algorithm and expansion algorithm to eliminate the noise in picture, improve Accuracy of detection.

Accompanying drawing explanation

Fig. 1 is the flow chart of the present invention；

Fig. 2 is to tear the detection schematic diagram in glove stub area；

Fig. 3 is to tear the detection schematic diagram in glove stub area；

Figure is designated: 1 industrial camera, 2 glove, 3 tear-off portions, 4 black backgrounds.

Detailed description of the invention

The present invention is described in detail with specific embodiment below in conjunction with the accompanying drawings.

As it is shown in figure 1, a kind of glove based on machine vision tear and greasy dirt detection method, for detecting on production line Glove 2, comprise the following steps:

(1) industrial camera 1 is used to shoot several continuous pictures of multiple glove 2；In order to improve detection quality and detection speed Degree, in conjunction with industrial practical situation, adds black background 4 on a production line during shooting.Due to glove 2 in commercial production The rotation of self it is accompanied by, it is contemplated that the detection for glove 2 to be seen from different perspectives during translation Examining the picture of glove 2, the present invention uses a fixing industrial camera 1 to shoot several consecutive images, utilizes glove 2 at production line On self rotate and obtain for the multi-angle image of glove 2.The visual field of one industrial camera 1 exists multiple glove 2, Therefore, it is necessary to same glove 2 are carried out visual tracking in multiple image.

(2) utilize Kalman filter that glove 2 are carried out visual tracking, obtain the figure of multiple angles of same glove 2 As information；Particularly as follows: the entirety of glove 2 and fingerprint to be regarded as a rectangle, take this rectangular centre as Kalman filter Input, estimates the position of these glove 2 in next frame image according to this input, obtains the image of multiple angles of same glove 2 Information, it is achieved tracing task.Because Kalman filter can by when in previous frame image the location estimation of glove 2 go out next The position of glove 2 in two field picture, compares the method that in direct comparison two frame picture, the change of glove 2 position is tracked, permissible Obtain and follow the tracks of result more accurately and have more preferable robustness for noise.

(3) glove 2 are carried out contours extract, and be cut into suitable size；When having obtained same glove 2 in multiple image Positional information after, cut out the picture comprising only glove 2, and multiframe picture combined, obtain the same hand set 2 The image information of multiple angles.But, before starting to detect glove 2 defect, in order to improve the speed of defects detection, protecting In the case of card photo resolution is certain so that the size of picture is the least, just containing the picture of glove 2 is to manage most Think.In order to realize such purpose, the present invention uses Canny edge detection operator, glove 2 is carried out contours extract, finds In profile the highest, minimum, the most left, four the rightest points construct a rectangle, the scope cutting covered by this rectangle Out, as next step picture to be processed

(4) whether detection glove 2 tear；The color particularly as follows: whether (401) judge to have powerful connections in the picture of glove 2 profile, If it has, then glove 2 tear, otherwise go to step (402)；

(402) judge in addition to glove 2 end, if having the color of glove layer material superposition background colour, if it has, then hands Set 2 tears, otherwise, glove 2 without tearing, described glove stub area with glove edge as partial contour, and only one layer The color of glove material superposition background colour.

Being usually present some noises, such as camera lens during detection profile the cleanest etc., this will cause detecting To profile in addition to the outline of glove 2 fingerprint and the Internal periphery that tears, also have some the least little profiles, in order to arrange Except the interference of these noises, before picture is carried out contours extract operation, erosion algorithm and expansion algorithm is used to eliminate these and make an uproar Point.

For the detection torn based on profile, because when glove 2 are deposited in case of tearing, being equivalent to more qualified hands How set 2 has compared an Internal periphery.Therefore when on the basis of outline, if detecting unnecessary profile, then these glove 2 may Exist and tear this defect.It is folded that what industrial camera 1 was collected can be regarded as two-layer glove material without glove 2 picture torn It is added on black background 4 a kind of color effects showed, but, at the end of glove 2, due to industrial camera 1 Setting angle and glove 2 deformation on a production line so that the image of this partial glove 2 is that glove layer material is superimposed upon background On, can be more deeper compared with color for other parts.Due to the existence of this part, tearing of glove 2 can be divided into two by us Class: the first is the tear-off portion 3 end section (as shown in Figure 2) at glove 2 of glove 2, this make tear-off portion 3 directly in Existing black background 4 color, it is easy to it is detected；The tear-off portion 3 of the second glove 2 is not at the end section of glove 2 (as shown in Figure 3), this makes the tear-off portion 3 of glove 2 also be that glove layer material is superimposed upon on background, and the end of glove 2 End portion has the same color performance, can be made a distinction it by the distribution of two parts profile coordinate, to get rid of glove 2 End section tears the interference of detection for the second, because the contoured profile of glove 2 end section is sufficiently close to image base, There is a certain distance in profile coordinate and image base that the second tears, otherwise this will be the first situation about tearing.

(5) whether detection glove 2 have greasy dirt.Owing to greasy dirt typically exhibits faint yellow, perusal is not very obvious, It is easy to during detection be ignored.The present invention conversion by color space so that greasy dirt is presented on HSV color space H dimension, and the white of the black of background and glove 2 is presented on V dimension, and rgb color space is by greasy dirt, glove 2 and background Color is coupling in three dimensions and compares, and has reached uncoupled purpose, simultaneously as greasy dirt has higher compared to glove 2 Reflective, so assisting the parameter with S space to arrange in HSV space, can preferably distinguish greasy dirt

Step (5) particularly as follows:

(501) according to formula, the color space of picture is transformed into HSV from RGB,

$s=\left\{\begin{array}{cc}0,& \begin{array}{cc}if& max=0\end{array}\\ \frac{max-min}{max}=1-\frac{min}{max},& otherwise\end{array}\right.$

V=max

Wherein, r, g, b are three parameter transform in artwork rgb color space to the value in [0,1] interval, max respectively Being r, the maximum in g, b, min is r, the minima in g, b, and h represents that tone, s represent that saturation, v represent lightness；

(502) if 40≤h≤50, then having greasy dirt on glove, otherwise, glove are without greasy dirt.

Claims (7)

1. glove based on machine vision tear and a greasy dirt detection method, for detecting the glove on production line, and its feature It is, comprises the following steps:

(1) industrial camera is used to shoot several continuous pictures of multiple glove；

(2) utilize Kalman filter opponent set to carry out visual tracking, obtain the image information of multiple angles of same glove；

(3) opponent packs into row contours extract, and is cut into suitable size；

(4) whether detection glove tear；

(5) whether detection glove have greasy dirt.

A kind of glove based on machine vision the most according to claim 1 tear and greasy dirt detection method, it is characterised in that In described step (1), during shooting, increase black background on a production line.

A kind of glove based on machine vision the most according to claim 1 tear and greasy dirt detection method, it is characterised in that Described step (2), particularly as follows: the entirety of glove and fingerprint is regarded as a rectangle, takes this rectangular centre as Kalman filtering The input of device, estimates the position of these glove in next frame image according to this input, obtains the figure of multiple angles of same glove As information, it is achieved tracing task.

A kind of glove based on machine vision the most according to claim 1 tear and greasy dirt detection method, it is characterised in that Described step (3), particularly as follows: use Canny edge detection operator, finds the highest, minimum, the most left and the rightest of glove profile Four points, one rectangle of structure, the scope cutting that this rectangle is covered out, as next step picture to be processed.

A kind of glove based on machine vision the most according to claim 1 tear and greasy dirt detection method, it is characterised in that Described step (4) particularly as follows:

(401) whether color, if it has, then glove tear, otherwise go to step if judging to have powerful connections in the picture of glove profile (402)；

(402) judge in addition to glove stub area, if having the color of glove layer material superposition background colour, if it has, then hands Being cased with tearing, otherwise, glove are without tearing, and described glove stub area is with glove edge as partial contour, and only one layer of hands The color of cover material superposition background colour.

A kind of glove based on machine vision the most according to claim 1 tear and greasy dirt detection method, it is characterised in that Described step (5) particularly as follows:

(501) according to formula, the color space of picture is transformed into HSV from RGB,

$s=\left\{\begin{array}{cc}0,& ifmax=0\\ \frac{max-min}{max}=1-\frac{min}{\max },& otherwise\end{array}\right.$

V=max

Wherein, r, g, b are three parameter transform in artwork rgb color space to the value in [0,1] interval respectively, and max is r, Maximum in g, b, min is r, the minima in g, b, and h represents that tone, s represent that saturation, v represent lightness；

(502) if 40≤h≤50, then having greasy dirt on glove, otherwise, glove are without greasy dirt.

A kind of glove based on machine vision the most according to claim 1 tear and greasy dirt detection method, it is characterised in that Before opponent packs into row contours extract in described step (3), erosion algorithm and expansion algorithm is used to eliminate the noise in picture.