CN208672539U - A kind of foliated glass edge faults detection device based on Image Acquisition - Google Patents

Abstract

A kind of foliated glass edge faults detection device based on Image Acquisition, comprising: roller frame, several transfer rollers, several detection cameras and processor；The pivot center of transfer roller is arranged in parallel so that the foliated glass in transfer passages can be supported on a transport plane；Camera is detected to be fixedly installed relative to roller frame so as to detect the camera lens of camera towards transfer passages；Detection camera is set to the outside of transfer passages and at least there are two the heteropleurals that detection camera is set to transfer passages；Processor and detection camera are constituted and are electrically connected.Having the beneficial effect that for the utility model provides a kind of optical detection apparatus suitable for industrial production line, which still is able to get a distinct image, and the clear image is used for industrial detection without adjustment object under test/camera spatial position in advance.

Description

A kind of foliated glass edge faults detection device based on Image Acquisition

Technical field

The utility model relates to industry detection apparatus.

Background technique

The precision high to the testing requirements of industrial products surface blemish.When using artificial detection industrial products surface blemish, It needs to go through industrial products surface by the eyes of detection workman, it is time-consuming and laborious.And the precision detected relies primarily on work The uncertain factors such as experience, eyesight, the physical strength of people can not be suitable for the industrial production of high-precision requirement.

Meanwhile when through the image of the image acquisition devices industrial products such as camera, video camera, root is generally required According to optical imagery relationship, parameters or the cameras, video camera and industrial products such as camera, the focal length of video camera, aperture are adjusted Between the relativeness such as direction, distance.Such set-up procedure makes the work for acquiring industrial products image become complex.

It is special due to the limitation of camera field depth itself although present camera has the function of autozoom energy It is not in macroshot, camera can not be operated by adjustings such as autozooms and obtain the complete of the body surface with certain length Whole, clearly image.

Utility model content

A kind of foliated glass edge faults detection device based on Image Acquisition, comprising: if roller frame, several transfer rollers, Dry detection camera and processor；Transfer roller is rotatably connected to roller frame to constitute one and be used to transmit with one fixed width The transfer passages of foliated glass；The pivot center of transfer roller is arranged in parallel so that the foliated glass in transfer passages can be supported on On one transport plane；Camera is detected to be fixedly installed relative to roller frame so as to detect the camera lens of camera towards transfer passages；Detection Camera is set to the outside of transfer passages and at least there are two the heteropleurals that detection camera is set to transfer passages；Detect camera lens The optical axis of lens is parallel to transport plane and the optical axis of at least two camera lenses for being located at the ipsilateral detection camera of transfer passages is uneven Row setting；Processor and detection camera are constituted and are electrically connected.

Further, the glass flaws detection device based on Image Acquisition further include: motor, motor are fixedly attached to roller Frame, the electric machine shaft driving of motor are connected to one or several to drive transfer roller to rotate in transfer roller.

Further, the glass flaws detection device based on Image Acquisition further include: photoelectric sensor, photoelectric sensor packet Include: light emitting head and light receiver head, wherein it is opposite to be separately positioned on transfer passages for light emitting head and light receiver head Two sides.

Further, the line between light emitting head and light receiver head is parallel to the pivot center of transfer roller.

Further, the updrift side of the transfer passages of transfer roller composition is arranged in light emitting head and light receiver head； The downstream direction of the transfer passages of transfer roller composition is arranged in detection camera.

Further, the foliated glass edge faults detection device based on Image Acquisition further include: Front camera, setting exist The top of transport plane, the camera lens of Front camera is towards transfer roller, and the primary optical axis of the camera lens of Front camera and transport plane hang down Directly.

Further, Front camera is line-scan digital camera, and line-scan digital camera includes several photosensitive elements, photosensitive element substantially edge It is parallel to the direction setting of transfer roller.

Further, the glass flaws detection device based on Image Acquisition further include: photoelectric sensor, photoelectric sensor packet Include: light emitting head and light receiver head, wherein it is opposite to be separately positioned on transfer passages for light emitting head and light receiver head Two sides；The updrift side of the transfer passages of transfer roller composition is arranged in light emitting head and light receiver head；Detect camera setting In the downstream direction for the transfer passages that transfer roller is constituted；Front camera is arranged between photoelectric sensor and detection camera.

Further, lighting device is arranged in above the transmission plane of transfer roller composition.

Further, the detection camera positioned at the transfer passages two sides that transfer roller is constituted is symmetrical about the split of transfer roller Setting.

Having the beneficial effect that for the utility model provides a kind of optical detection apparatus suitable for industrial production line, the dress It sets without adjustment object under test/camera spatial position in advance, still is able to get a distinct image, and the clear image is used for work Industry detection.

Detailed description of the invention

Figure 1A -1C is a variety of image capturing system schematic diagrames of the utility model；

Fig. 2 is scanning camera working state schematic representation；

Fig. 3 A is a kind of system block diagram of industrial detection system；

Fig. 3 B is a kind of schematic diagram of industrial detection system；

Fig. 4 is a kind of encoder mounting means schematic diagram；

Fig. 5 A is a kind of stereoscopic schematic diagram of camera constituted with biasing；

Fig. 5 B is a kind of schematic internal view of camera constituted with biasing；

Fig. 6 A is the schematic diagram of the image acquisition subsystem comprising 3 cameras；

Fig. 6 B-6D is the positional relationship of the blur-free imaging face P of each camera of the image acquisition subsystem comprising 3 cameras Schematic diagram；

Fig. 7 A is a kind of schematic diagram of edge ground glass flaw parts of images；

Fig. 7 B is the schematic diagram of another edge ground glass flaw parts of images；

Fig. 8 A is the schematic diagram of the strip-shaped light source of industrial detection system；

Fig. 8 B-8D is the schematic diagram of the cyclic annular light source of industrial detection system；

Fig. 9 A is to know that the industrial detection system of system judges edge ground glass using the image before fusion comprising position The flow chart of edge faults；

Fig. 9 B is to know that the industrial detection system of system judges edge ground glass using fused image comprising position The flow chart of edge faults.

Specific embodiment

Blur-free imaging face P

For an image collecting device, such as camera, video camera, scanner etc., by optical principle it is found that if not having There are physical conditions, the camera such as optics, the model of electronic component, position in change image collecting device to be capable of the distance of blur-free imaging It is often fixed with range.

As shown in figs. 1A-1 c, the region that will be clearly shot by camera 101 is defined as blur-free imaging face P.In general, The shape of blur-free imaging face P depends on the hardware of camera 101, the especially shape of imaging sensor, in Figure 1A -1C just to Illustrate and its example is drawn as rectangle, in fact, blur-free imaging face P can be arbitrary shape.The size of blur-free imaging face P and Its between camera 101 at a distance from, angle, determined by the relativeness between lens and imaging sensor.

Image capturing system

Image capturing system includes at least image collecting device and processor, here using camera as image collecting device.

The blur-free imaging face P of camera is parallel with target object surface

When carrying out macro, generally requires camera and closely shoot subject 104.

As shown in Figure 1A, 102 be carrying apparatus, is used for drop target object 104.103 be locating piece, for fixing target The position of object 104.If it is precisely plane that subject 104, which needs the part shot, and just in blur-free imaging plane P It is interior, it can just obtain partially complete, the clearly image that is taken.

If but as shown in Figure 1B, 104 position of subject changes (or planar structure is not present in itself), this If when subject 104 shot by camera 101, since the depth of field of the camera of shooting at close range is smaller, thus subject The 104 not part in the P of blur-free imaging face then can not be clearly imaged.

In this case, even if by adjusting the focal length of camera 101 or the phase of adjustment camera 101 and subject 104 It adjusts the distance, can not obtain subject 104 and be taken partially complete, clearly image.

The blur-free imaging face P and target object surface of camera are not parallel

The utility model provide a kind of image capturing system include: camera 101, driving equipment 107 and processor (Figure 1A and Figure 1B is not shown).

Wherein, camera 101 is used to shoot the image of a subject.Camera 101 can use varifocal camera can also Using the camera of fixed focal length.

Driving equipment 107 can both drive camera 101 or drive subject 104, or drive them simultaneously.This In embodiment, driving equipment 107 is a motor.As a kind of specific scheme, driving equipment includes one for contacting, containing The carrying apparatus 102 of subject 104 is put or transmits, which is driven the driving of equipment and drives and be taken The movement of object 104.Such benefit is, it is not necessary to continually adjust camera 101.

As shown in Figure 1B, 1C, in the case where the blur-free imaging face P of camera and body surface are not parallel, adjust in any case Focal length/object distance is saved, the blur-free imaging face P of camera cannot be completely be overlapped with the surface of target object.At this point, camera is clear Imaging surface P and target object surface often only have an intersection from geometrical relationship.In this case, the photo of camera shooting In only parts of images be that clearly, which is the friendship between the blur-free imaging face P of camera and target object surface Line.

Below for acquiring the image of a cube face, the surface of the cube and the blur-free imaging face P of camera It is not parallel, and think that the blur-free imaging face P of camera intersects with the surface of target object.

The clear part in image is identified by the peak value of image parameter gradient

As an alternative embodiment, the step of acquiring and judging clear part in cube graph picture is as follows:

Acquire cube face original image.In the image, in place of the intersection of the blur-free imaging face P and body surface of camera Image be that clearly, the image of other parts is fuzzy.The image acquired in the step can be gray scale or color image.

Denoising is carried out to cube face original image, obtains denoising image.Denoising can reduce due to hard Influence of the noise that part or environment generate to following image processing work.

Calculate the absolute value of the parameter value gradient in denoising image between adjacent pixel/pixel group.Here parameter can be with It is gray value, brightness value, is also possible to the other parameters such as contrast, saturation degree.

The peak point of gradient is judged using Gaussian function according to the absolute value for the gradient value that abovementioned steps are calculated x_max.In same picture, the pixel of " fuzzy " is considered as having taken the average value of the pixel around the fuzzy pixel, reduces The comparison of the pixel and neighboring pixel, to generate fuzzy visual effect.And between the pixel and surrounding pixel of " clear " Comparison it is then distincter, to more be able to reflect the details of pixel, visual effect is more clear.The influence for excluding noise, In same picture, gradient absolute value is bigger, then the image at this is also more clear.Alternatively, judge gradient value The mode of peak point can be other calculation methods in addition to Gaussian function.Peak point X_maxWhereabouts, the as picture In clear part.

During calculating gradient value, gradient value can be pressed the location of in picture according to pixel/pixel group Sequence arranges, that is, the calculated result of each gradient value corresponds to a certain specific location in picture.At this point, obtaining peak It is worth point X_maxWhile, can also obtain the peak point the location of in picture, that is, in the width image clear part tool Body position.

This peak value judgement can be realized using the gradient processing unit in processor.

The method that clear part is judged by image parameter peak value can be obtained in shooting surface texture object abundant More preferably effect.

By calculating the clear part in intersection point identification image

Establish coordinate system

As an alternative embodiment, by the line segment for the blur-free imaging face P for indicating camera and target object surface wheel Exterior feature indicates in the same coordinate system.The coordinate system can be two-dimensional coordinate system, coordinate plane perpendicular to camera it is clear at Image planes P.

The position of fixed camera, and the parameter constant of the lens of camera and imaging sensor.At this point, camera it is clear at The position of the corresponding line segment p of image planes P in a coordinate system is also just fixed.Selected x-axis, y-axis and coordinate origin, then it is every on line segment p One point has fixed coordinate value.Line segment p can be indicated by the linear function in certain section.Line segment p and target Position where the intersection point of profiling object surface corresponds to the position of the clear part in image.

Obtain the coordinate information of target object surface profile

As shown in Fig. 2, obtaining the coordinate information of profiling object surface using scanning camera 201.Scanning camera is arranged in mesh The top of object is marked, usually surface.Movement mechanism makes to generate relative motion between scanning camera and target object, so as to sweep Body surface can be traversed everywhere by retouching camera.In this schematic diagram, scanning camera 201 is mounted on guide rail 202 by movement mechanism. The shooting direction of scanning camera is perpendicular to coordinate plane, to obtain the projected outline of target object in a coordinate system.In the two In relative movement, the distance between scanning camera 201 and plane where coordinate system are kept constant.Scanning camera 201 can be with Using line-scan digital camera or area array cameras.Direction of relative movement between camera and target object is defined as the direction x, perpendicular to the x Direction is the direction y.

Line-scan digital camera obtains profiling object surface coordinate information

Line-scan digital camera is the camera using line scan image sensor.The image obtained compared to area array cameras, line-scan digital camera Resolution ratio it is higher, but since its primary shooting can only obtain a width row image, the figure of complete object to be shot in order to obtain As needing by subsequent processing operation.

The coordinate information of profiling object surface is obtained using line-scan digital camera, specific work step is as follows:

Line-scan digital camera scanning: the motion range of line-scan digital camera is fixed, and in fixed spacing distance continuous scanning, and is exported Image, each image are a line.

Threshold process: by image that line-scan digital camera obtains and it is unobstructed when Background make the difference, by the exhausted of difference or difference Value is compared with preset threshold value, when difference in a certain range, judge line-scan digital camera obtain the row image do not include Target object, at this point, the gray value of image at this is adjusted to a certain fixed value.When difference is more than a certain range, then it is assumed that this The image at place is the image of target object, at this point, the gray value of image at this is adjusted to another fixed value and by line-scan digital camera Two endpoints of the target object image (row image) of acquisition are recorded.Operation in this way can obtain object and cover The image in the region of lid, and the image is bianry image.Bianry image is more convenient for handling.And image information only records row figure Two endpoints of picture, are used to form contour of object, have given up other useless information, have reduced the memory capacity of system.

The direction x coordinate obtains: movement mechanism makes line-scan digital camera and target object generate relative motion, utilizes encoder equipotential The displacement information that acquisition device obtains any time movement mechanism is moved, the displacement information of the moment target object is also just obtained. The encoder output corresponding displacement information of row image.Since the shooting direction of line-scan digital camera is perpendicular to coordinate plane, The location information of the corresponding encoder of row image is the x coordinate information of the row image.

The direction y coordinate obtains: the length of line-scan digital camera scan line is fixed, and the y-coordinate of one end of scan line is set as 0, then The y-coordinate of two endpoints all passes through the distance between to calculate the two endpoints and scan line endpoints and acquire in the row image.

Smoothing processing: by line-scan digital camera obtain front and back multiple image splice, and to spliced profile information into Row smothing filtering obtains the calibration curve information at whole object edge.Smothing filtering can eliminate single-frame images processing bring error, So that the profile of target object is reflected true information more accurately, is conducive to subsequent processing.

Area array cameras obtains profiling object surface coordinate information

Area array cameras uses array image sensor.Its imaging region is a face, can be obtained by once shooting The image of complete object to be shot.

The coordinate information of profiling object surface is obtained using area array cameras, specific work step is as follows:

Area array cameras shooting: target object is placed in the coverage of area array cameras, to ensure that area array cameras is once clapped Take the photograph the complete image that can obtain target object.If the incomplete image of image of area array cameras acquisition, can refer to linear array at this time The complete image of the work step acquisition target object of camera.

Threshold process: line-scan digital camera is similar to the step of threshold process.

X, the direction y coordinate obtains: area array cameras position is fixed, and the range of shooting is also fixed.Define area array cameras shooting The coordinate of certain boundary point of range is (0,0), then can be passed through by the x for the profile point that bianry image indicates, y-coordinate information in terms of The distance between the profile point and the boundary point is calculated to obtain.

Smoothing processing: aforementioned line-scan digital camera is similar to the step of smoothing processing.

Intersection point is calculated to obtain the clear position in image

As an alternative embodiment, the step of acquiring and judging clear part in cube graph picture is as follows:

Acquire the original image of the cube face.Acquired image can be gray level image, be also possible to cromogram Picture.

Scanning camera scans the image of target object, obtains the coordinate information of target object edge contour.Camera it is clear The coordinate information for the line segment p that imaging surface P is projected in a coordinate system is demarcated in advance.

Calculate the intersection point between line segment p and the edge contour of target object.The image of the point of intersection is in original image Clear part.At this point, processor has known relative position information of the blur-free imaging part in target object.

Using laser labelling blur-free imaging face to identify the clear part in image

In the case where the blur-free imaging face of camera and target object surface are not parallel, if in the image of camera acquisition at this time There are still clearly parts, then the blur-free imaging face P of camera can be approximate from geometrical relationship with the surface of target object Think that there are an intersections in ground.

Using the position in laser labelling blur-free imaging face, then any time target object surface is by the place of illuminated with laser light For the intersection on the surface of the blur-free imaging face P and target object of camera.When being reflected in the image of camera shooting, quilt in image The place of illuminated with laser light is the position in the image where clear part.

As shown in Figure 1B, as an implementation, laser beam emitting device 108 is arranged in the blur-free imaging face P's of camera Side, including several laser emitters, the laser beam that several laser transmitter projects go out is parallel to each other, Shu Jiguang each In on same plane, forming laser region, laser region covers the blur-free imaging face P of camera.As another optional implementation The top of blur-free imaging face P is arranged in mode, as shown in Figure 1 C, laser beam emitting device 108, and laser emitter emits from top to bottom Planar laser is that target object surface is overlapped with laser region partially due to surface itself since laser has certain width Diffusing reflection, still will appear laser facula on the surface of object without being obscured by an object completely.Planar laser region covers The blur-free imaging face P of lid camera.In the present embodiment, laser uses wavelength for the red laser of 650nm.It is optional real as other Mode is applied, the color of laser is also possible to green or other colors.

As a kind of mode for judging clear part in image according to laser facula, camera uses polychrome camera, including the One channel and second channel.In the present embodiment, camera uses color camera, including tri- channels RGB, in the present embodiment, laser Use wavelength for the red laser of 650nm, first passage is the channel R, and second channel is the channel G or channel B.It is worth mentioning It is the first passage defined here, second channel can include several channels, such as, second channel It can be the channel GB.Processor includes channel selection unit, for selecting the channel of polychrome camera.Processor extracts polychrome camera The first passage image of acquisition judges the laser facula region in first passage image.Then second channel image is extracted, and will Laser facula region is matched with second channel image in first passage image.Laser facula region is in first passage image The region of clear image in second channel image.As optional embodiment, when the color of laser is green, first passage For the channel G.Laser facula region and clear image position are judged using polychrome camera respectively, can be avoided the irradiation of laser to figure As the interference of information, facilitates and carry out the subsequent operations such as Defect Detection using image.

Keep the light distribution of laser region uniform using optical elements such as gratings as optional embodiment.Uniformly Laser intensity distribution will not bring excessive interference to target object surface information.In this case, without polychrome phase Machine filters laser.

By the way of laser labelling blur-free imaging face, different photographed scenes equally can adapt to, and eliminate scanning The equipment such as camera.

Image co-registration

The complete image of body surface in order to obtain, using processor by the corresponding clear figure of the different location of body surface As fusion, the clear image of complete body surface is formed.Before carrying out image co-registration, processor according to the systems and methods, Relative positional relationship of the part of each blur-free imaging relative to the part of other blur-free imagings in same image is obtained.

As an implementation, camera is mounted in driving equipment, can change camera site with driving equipment.Make For a kind of embodiment, driving equipment includes mounting base, stepper motor, transmission mechanism, wheel, guide rail.Camera and mounting base are solid Fixed connection.Stepper motor can be set inside mounting base, also may be mounted at the outside of mounting base.One end of transmission mechanism with Wheel is connected, and the other end is connected with stepper motor.Stepper motor drive transmission device is to drive wheel to rotate.Wheel can edge Guide rail movement, and its motion profile is determined by guide rail.In a shooting process, driving equipment according to guide rail motion profile, It is moved along certain direction, under the driving of stepper motor, camera changes identical distance every time.Camera is according to driving equipment Walking sequence, in the picture of different location shooting body surface.Processor receives these pictures, by judge gradient peak or The mode that person calculates intersection point obtains the clear part in these pictures, and in sequence by clearly part is spelled in these pictures It connects.(D in attached drawing is the direction of relative movement for indicating object and camera)

As optional embodiment, the motor of driving equipment can use servo motor.

As another embodiment, driving equipment can drive target object or camera to rotate around a rotation axis, Rotation axis is parallel to the blur-free imaging face P of camera.In this manner, target object surface different piece can also be obtained Image.

The position of alternatively optional embodiment, camera is fixed, and driving equipment drives the object fortune being taken It is dynamic.Camera shoots plurality of pictures in the different positions of target object.Processor receives these pictures, by judging gradient peak Value or the mode for calculating intersection point obtain the clear part in these pictures, and in sequence by clearly part in these pictures Splicing.At this point, driving equipment can be industrial producing line, roller bearing, transfer cart etc..

In short, driving equipment, for driving camera or/and target object, so that the part to be captured of target object can be divided When appear in the blur-free imaging face P of camera.As optional embodiment, image capturing system can also include positioning dress Set, constitute data connection with driving device and processor respectively so that processor can by image it is clear partially and institute's object The corresponding part of body is matched.

As optional embodiment, camera can be connected with processor.Camera sends the picture taken to processing Device, processor handles image, to obtain the information for the body surface for including in image.The information of body surface includes body surface Flaw information, text, flag information, the color of body surface, texture, the pattern of body surface etc..Wherein, flaw information is extremely It less include: the information that whether there is flaw in image；The location information of flaw in image.

Position in place of being intersected by calculating blur-free imaging face with body surface judges the clear part in image, Neng Goushi Various photographed scenes are answered, not will receive the influence of illumination condition or article surface vein.

Multiple cameras are shot jointly

It may include several cameras in image capturing system, to improve the efficiency of shooting.For example it is driven using rotary When dynamic equipment 107 is rotated with animal body, if only one camera, rotary driving equipment needs to be rotated by 360 ° under normal conditions The complete image of target object surface can be collected.The camera being oppositely arranged according to two, target object are placed on two Between a camera, the blur-free imaging face of two cameras is perpendicular to same plane.At this point, driving equipment only needs to rotate 180 ° of energy It is enough to complete shooting, reduce the time of half.According to three cameras, centerlines between any two are 120 °, two two-phases Plane where the blur-free imaging face of machine tilts intersection.At this point, driving equipment, which only needs to rotate 120 °, can complete to shoot.

On-line detecting system

Industrial detection often examination criteria with higher.If replacing visual inspection with Image Acquisition detection, The shooting of camera short distance object to be detected is needed to obtain finer body surface image information.

As shown in Fig. 3 A, 3B, a kind of industrial detection system based on Image Acquisition, including industrial producing line 30, position are known Subsystem 31, image acquisition subsystem 32, processor 33.

Below by taking glass edge on-line detecting system as an example, specific image capturing system of the introduction comprising the utility model On-line detecting system, in the present embodiment, glass 39 to be measured is foliated glass.As optional embodiment, this detection system The surface blemish that can detecte other objects in addition to glass, such as wood surface flaw, steel surface flaw, the stone surface flaw Defect etc..

Wherein, subsystem 31 is known for obtaining the coordinate information of glass edge to be measured, by calculating image ginseng in position Number gradient peaks or using laser labelling blur-free imaging face to identify the clear part in image when, position knows that subsystem can be with It omits.Image acquisition subsystem 32 includes one or more cameras, for acquiring the image of glass edge.Processor 33 at least with Position knows that subsystem 31, image acquisition subsystem 32 constitute data connection, for carrying out each item of image and data processing operation.

The industrial detection system can obtain the clear image on object under test surface by machine.Save artificial, raising Detection accuracy.Using this industrial detection system, determinand can be put in industrial producing line with arbitrary position, direction.

Industrial producing line

Industrial producing line 30 is for placing the glass being taken.Industrial production line at least can by object to be detected one transport face with One, which transports direction, transports straight line along one and is transported.In the present embodiment, industrial producing line drives the glass being placed in industrial producing line Glass movement.

Industrial producing line includes sending unit, for transporting product to be measured.

As an alternative embodiment, industrial producing line includes roller frame and several transfer rollers, several transfer rollers with Roller frame rotatably connects, and the axis of several transfer rollers is in the same plane.

As other optional embodiments, industrial producing line also may include the feeding devices such as conveyer belt, delivery vehicle.

The surface of industrial producing line can use anti-slip rubber material, or setting sucker, for increasing and being transported object Between frictional force improve the transport efficiency of industrial producing line to increase the frictional force between industrial producing line and glass.

Know subsystem 31 in position

Position know subsystem for obtaining the coordinate information of product to be measured, by calculate image parameter gradient peak or When using laser labelling blur-free imaging face to identify the clear part in image, position knows that subsystem can be omitted.

Face is transported as the plane where coordinate system using industrial producing line, establishes coordinate system.The coordinate system can be sat for two dimension Mark system.The contour of object placed in industrial producing line projects into a closed figure in the coordinate system.With the defeated of industrial producing line Fortune direction is the direction x, and transporting direction perpendicular to this is the direction y.Position know subsystem at least need to know product to be detected with Relative position between camera.

Front camera

As shown in Figure 3B, the top of industrial producing line 30 is arranged in Front camera 312, is being generally arranged at industrial producing line 30 just Top.Certainly, as optional embodiment, Front camera 312 can be configured according to actual needs.But it sets in any case It sets, will guarantee that the areas imaging of Front camera 312 can cover suitable industrial producing line region.Front camera 312 and industry Producing line 30 is separated by a distance.The industrial producing line 30 of camera lens direction of Front camera, and the axis of its optical axis and several transfer rollers The plane at place is vertical.

Front camera can be line scan camera (line-scan digital camera) or area array cameras.Scanning camera is to be measured for obtaining The coordinate information of object.

As optional embodiment, Front camera 312 is line scan camera.

The acquisition of industrial producing line location information (glass contours x coordinate value)

Glass to be measured is placed in industrial producing line, and as industrial producing line moves together.From coordinate plane, any Moment, as long as obtaining the location information of industrial producing line, it will be able to obtain glass to be measured in the x coordinate value at the moment.

As an implementation, to Front camera obtain image judge, by same glass to be measured at first by The x coordinate value of the collected part of Front camera is set as 0.

As optional embodiment, the direction y of industrial producing line is provided with a position detecting device.As a kind of embodiment party Formula, which is photoelectric door, including being arranged in the sender unit of industrial producing line side and being oppositely arranged The line of signal receiving device, sender unit and reception device transports straight line perpendicular to industrial producing line.It will be on the line The x coordinate at any point is set as 0.In the case where not blocking, the reception device of photoelectric door can receive photoelectric door always Light emitting device issue optical signal.In the case where blocking, the reception device of photoelectric door cannot receive the hair of photoelectric door The optical signal that electro-optical device issues.

Photoelectric door is connected with processor, and photoelectric door is kept constant at a distance from Front camera.At this point, Front camera scans The x coordinate information in region can also obtain.

Encoder obtains coordinate

As a kind of displacement detector, encoder includes coding disk and coding reading unit, for angular displacement or directly Displacement of the lines is converted into electric signal.

As shown in Fig. 3 B, Fig. 4, the industrial detection system based on Image Acquisition includes encoder 311, and encoder can be rotation Turn encoder, including absolute value encoder and incremental encoder.

The coding disk of encoder can be connected to the motor by shaft coupling, can also with for direct industrial producing line movement Transmission device is connected.When encoder is connected to the motor, what encoder directly obtained is the rotation information of motor, is needed by subtracting Speed ratio converts to obtain the coordinates of motion of corresponding industrial producing line.

Encoder is connected with processor, for sending information to processor.The image that is obtained for the first time according to scanning camera or Photoelectric door is set out, and the coordinates of motion of the encoder industrial producing line of encoder record at any time, are exactly that the x put before glass is sat Mark.

Location information due to obtaining industrial producing line is the original steps and key step of entire detection system detection process Suddenly.If the step for produce mistake, the output result of that subsequent various step is likely to generate mistake.Therefore it needs Ensure the accuracy for the information that encoder obtains.

As the scheme of substitution photoelectric door, processor calculates the real-time speed of industrial producing line according to the data of encoder transmission Degree.At the time of the real-time speed of industrial producing line declines suddenly, judge to placed glass in industrial producing line.When by the speed rapid drawdown The coordinates of motion for carving industrial producing line are labeled as 0, at this point, the x coordinate put before glass namely 0.

As optional embodiment, glass edge detection system includes multiple encoders, and multiple encoders are mounted on work On the different location of industry producing line or different components.Such as using two encoders, main encoder 311a and pair Encoder 311b is connected with motor, industrial producing line transmission mechanism respectively.If information and main coding that pair encoder 311b is obtained The information that device 311a is obtained is consistent or error is within the scope of certain, then it is assumed that the information that encoder obtains at this time It is accurate.The setting of such major and minor encoder can find encoder fault in time, improve detection accuracy.

Coordinate is obtained using brushless motor parameter

Brushless motor is controlled by electronics commutation instruction, and speed can be controlled by pulse-width signal.Due to This characteristic of brushless motor enables brushless motor controller to obtain brushless electricity in such a way that record controls number of signals The rotational angle of machine, to obtain the motion information of industrial producing line.

When photoelectric sensor exports zero coordinate, glass enters detection zone, and brushless motor controller records brushless motor Control signal pulsewidth and number calculate the movement of industrial producing line at any time and according to the pulsewidth and number of control signal The x coordinate put before coordinate, that is, glass.

The time interval of Front camera shooting

Encoder or brushless motor are either used, the real-time speed of industrial producing line can be accessed.Since industry produces The load placed on line is different, or since the lubricant environment of industrial producing line mechanical device is distinguished, industrial producing line is frequently not even Speed movement.At this point, the real time kinematics speed of industrial producing line is calculated in processor.It calculates at such speeds, industrial producing line movement Time required for fixed intervals, and in the shooting of triggering of corresponding time point Front camera.It is shot using this Front camera The triggering mode at moment can guarantee that the picture of Front camera shooting meets the requirement of post-processing.For example it is adopted in Front camera In the case where line-scan digital camera, it can be ensured that the movement space that line-scan digital camera shoots corresponding industrial producing line every time is constant；Preceding Camera is set using in the case where area array cameras, it can be ensured that area array cameras is shot every time can access complete glass surface figure Picture.

Specific shooting interval is arranged when for using line-scan digital camera, is with the demand under the requirement of different image quality It is quasi-.Such as larger-size glass to be measured, the filming frequency relative reduction of line scan camera reduces energy consumption, reduces The fever of line scan camera.It is of course also possible to increase line scan camera filming frequency, to increase the fine degree of imaging.

After the completion of Front camera shooting, in conjunction with the corresponding industrial producing line coordinate of encoder or brushless motor control parameter Information can obtain the coordinate information at any time glass edge any point.

Image acquisition subsystem 32

Image acquisition subsystem acquire image, processor by judge image parameter gradient peak, calculating intersection location or The clear part in acquired image is identified in the way of laser labelling blur-free imaging face.In the present embodiment, Image Acquisition Subsystem 32 is used to obtain the image of glass edge to be measured.Image acquisition subsystem 32 may include multiple cameras 321.

As an implementation, image acquisition subsystem 32 is arranged after subsystem 31 is known in position, glass to be measured When being transported, first passes through position and know subsystem 31, using image acquisition subsystem 32.Such setting, enables to figure For picture acquisition subsystem 32 when acquiring image, the coordinate information of glass edge to be measured is it is known that facilitate subsequent processing operation.

The course of work of single camera in image acquisition subsystem

In the case where detecting glass edge flaw, it is often any that glass is placed on position, direction in industrial producing line 's.

As an implementation, as shown, the position of camera is fixed, the center of areas imaging and industrial producing line are protected Water holding is put down and the blur-free imaging face P of camera at least partly covers industrial producing line from the projection right above industrial producing line.It is produced in industry In the motion process of line, the position of camera is remained unchanged.

Glass to be measured is placed in industrial producing line with arbitrary direction.In this case, glass edge and camera at As often not parallel.In the glass edge image of camera shooting, the point of intersection of glass edge and camera blur-free imaging face P Image be that clearly, the other parts in image other than intersection point are unsharp.

As an alternative embodiment, processor knows the real-time position for taking glass to be measured that subsystem obtains according to position It sets, controls the shooting of camera.Soon enter camera shooting area in the preceding point of glass to be measured or comes into camera shooting area When domain, processor control camera starts to shoot.

In order to guarantee in the motion process of glass to be measured, the every bit of glass edge can be with image acquisition subsystem 32 some blur-free imaging face P intersection, as an alternative embodiment, position restrainer can be arranged at the edge of industrial producing line Structure is to prevent glass edge to be measured from exceeding the width range of industrial producing line；As another optional embodiment, it can introduce Warning device/function: when Front camera 312 judges that the shape of glass to be measured, location information exceed the width range of industrial producing line, Then alarm.

Shooting plurality of pictures simultaneously carries out image co-registration

Clear part in the photo that camera is shot there are the depth of field, camera has certain range.

In order to obtain the complete image of glass edge, camera is continuously shot multiple pictures, every image at the time of specific In it is clear part splice after be capable of forming the complete glass edge image of a width.

The horizontal direction central point in blur-free imaging region in the i-th frame image is denoted as x_i。

With x_iCentered on, the both horizontally and vertically range W of interception image_xAnd W_y.Wherein, x_i,W_xAnd W_yAll it is with pixel Unit.Definition range in the image of camera shooting are as follows: [x_i-W_x/ 2, x_i+W_x/2]。

The representation method of parameters includes: image distance in formula below: u, object distance: v, belt movement speed: V, object The mobile distance in vertical camera optical axis direction: S_v, as the mobile distance in vertical camera optical axis direction: S_u, as in sensor plane Abscissa: x (the pixel level position of the corresponding diagram on piece point), sensor pixel size: d, thickness of glass: T, camera lens are burnt Away from: f, sensor perturbations angle: θ, reference value x₀And u₀, give horizontal coordinate x on sensor₀, the image distance for corresponding to camera lens is u₀。

In the method for calculating image clearly part by image gradient peak point, need to calculate definition range in the image Region.Angle [alpha] of the movement speed V and camera optical axis of belt relative to belt；The exposure interval t of two field pictures.

In the method for calculating image clearly part by image gradient peak point,

W_x=S_u/ cos (θ)/d, W_y=(u*T)/(v*d)

Wherein, S_u=u*S_v/ v, v=1/ (1/f -1/u), u=u₀+Δu；S_v=V*sin (α) * t；Δ u=(x_i-x₀)* T is divided between the taking pictures of sin (θ) camera.

It is obtained in image in the algorithm of clear part by calculating intersection location,

W_x=S_u/ cos (θ)/d while W_y=(u*T)/(v*d)

Wherein, S_u=u*S_v/ v, v=1/ (1/f -1/u), u=u₀+Δu；S_v=V*sin (α) * t

In the method, the glass edge location information and encoder count that rearmounted camera is provided according to Front camera select Opportune moment takes pictures.

Processor controls image acquisition subsystem according to the relative position between the product and camera to be detected known Shoot work.The clear position x of image shot by camera_iOpportunity calculated especially by following methods:

Assuming that camera takes pictures and is divided into C between encoder twice for front and back_i-1+C_pbAnd C_i+C_pb, then glass position of taking pictures twice variation For (C_i-C_i-1)*S_c。

Because of camera sensor included angle θ, as plane x-axis direction Pixel Dimensions be d*Cos (θ), as plane y-axis direction Pixel Dimensions are still d.

The mobile distance S in the vertical camera optical axis direction of object_v=(C_i-C_i-1)*S_c*Sin(α)。

Further according to the location information (X for the marginal point that Front camera provides_i, Y_i+S_pb) and D, the object distance v of available camera =(D+X_i)/sin(α).Also image distance u can be obtained according to focal length of lens formula.

And then obtain x_i=(u-u₀)/sin(θ)+x₀。

For controller according to the above calculated result, binding site knows the information that subsystem obtains, and control camera is shot, so that The image arrived that camera is shot twice in interval time/distance, can satisfy the requirement of image mosaic.

Multiple clear images for corresponding to glass edge different location are merged to get into single camera coverage Complete glass edge image.

If current glass moving direction is right-to-left: [x-D/2, the x+D/2] image for acquiring present frame is placed on output The right side of image.

If current glass moving direction is from left to right: [x-D/2, the x+D/2] image for acquiring present frame is placed on output The left side of image.

Upper two step is repeated until obtaining complete clear image.

Sensor/lens tilt

In detection object surface blemish, object under test can be placed in industrial producing line with arbitrary direction.This is practical Novel Image Acquisition and detection system, by obtaining the clear figure of overlapping position between body surface and camera blur-free imaging face P Picture carries out the acquisition of body surface image.Image collecting device, including camera, video camera, scanner etc., blur-free imaging Face P has certain length.And this body surface Defect Detection system, the biggish field depth for needing camera to have, from And it is able to cover region sufficiently wide in industrial producing line.

It, can be using the method for replacement lens and/or imaging sensor in order to obtain biggish field depth.But it is bigger Field depth means more costs.

As an alternative embodiment, the image acquisition subsystem in industrial detection system, which uses, has biasing structure At camera.

As shown in figure 5, the camera that there is biasing to constitute includes shell 501, stationary lens 502 and imaging sensor 503, figure Picture sensor is connect with installation axle 504, and installation axle is rotatably attached to shell 501.Shaft portion is installed to expose on shell 501 Portion, rotation toggle 505 are fixedly connected with the part that installation axle exposes shell.As an alternative embodiment, rotation toggle Angled code-disc 506 is arranged in surrounding, for indicating the rotation angle of imaging sensor.The definition of lens 502 has a primary optical axis, schemes There is a sensing plane as sensor 503 defines, it is by installation axle 504 and flat perpendicular to the straight line of the primary optical axis of lens and sensing Face forms an angle α.Toggle 505 is operable is rotated for rotation, so that installation axle 504 be made to rotate, installation axle 504 is with motion video Sensor 503 rotates.By the angle code-disc 506 being arranged in around rotation toggle, present image biography can be intuitively known very much The tilt angle of sensor.

As optional embodiment, lens 502 can also rotate, to change inclining between imaging sensor and lens Oblique relationship.

For image taking sensor inclination, the focal range of sensor is indicated with following formula:

[(u+sin(α)*x)/(u+sin(α)*x-1),u/(u-1)]

Wherein, u is image distance, and α is that imaging sensor rotates angle, and x is imaging sensor in the corresponding focal length of rotary shaft radial direction Normalization size (x=X/f, wherein X be full-size(d)), sin (α) * x be sensor optical axis direction projector distance.

Corresponding given x and u it can be seen from formula, α bigger (u+sin (α) * x)/(u+sin (α) * x-1) is smaller, Mean that field depth is bigger.The range of α angle (can also state the primary optical axis and biography of lens as between 5 ° and 65 ° The range for the angle that sense plane is crossed to form is between 25 ° and 85 °) when, preferably imaging effect can be obtained.

According to the image-forming principle of lens, the camera that there is biasing to constitute, compared to using identical lens, image sensing The general camera of device, blur-free imaging face P have longer length, also correspond to bigger field depth.

The length and field depth of longer blur-free imaging face P enables the areas imaging of the camera constituted with biasing Cover region as much as possible in industrial producing line.Since the position that glass to be measured is put in industrial producing line is uncertain, using tool The camera for having biasing to constitute, can ensure as far as possible that any point in industrial producing line is being transported under identical hardware cost It can at a time generate and be overlapped with certain point on the blur-free imaging face P of camera in dynamic process, also ensure that glass Edge any position at a time can have biasing constitute camera in blur-free imaging.

Meanwhile bigger field depth can obtain more when shooting has the body surface image of certain depth Clear image.For example when shooting has the glass of radian, the camera that there is biasing to constitute, can under certain shooting angle Disposably shoot entire glass surface clearly microspur image.And traditional microspur camera then may be used since the depth of field is relatively small It can not can obtain simultaneously the clear image of the different location of the glass with radian.

With the camera that biasing is constituted there is biasing to constitute with the camera that biasing is constituted with the camera that biasing is constituted The camera that there is camera biasing to constitute

Multiple cameras are shot jointly

Object often has multiple surfaces, and there may be phases between the different location on multiple surfaces of object or surface Mutually block.At this point, single camera can not be in the image on the synchronization acquisition multiple surfaces of object.

Image acquisition subsystem 32 is used to acquire the complete image of glass edge.Image acquisition subsystem 32 includes multiple phases Machine.Multiple cameras can be set in same level.Under normal conditions, multiple cameras are distributed in the periphery of industrial producing line, and The two sides that industrial producing line transports straight line are set.The blur-free imaging face P of any two camera is not overlapped.

As shown in Figure 6A, as an alternative embodiment, multiple identical cameras are around industrial producing line width direction Certain point (point is known as central point) on middle line is uniformly distributed, what two lines between camera and central point of arbitrary neighborhood were formed Angle is consistent, and can also be described as multiple camera using central point to be uniformly distributed on the circumference of some circle in the center of circle. Such set-up mode keeps each camera in a center of symmetry according to the central point.Centrosymmetric relationship, so that entire figure As acquisition subsystem 32 installation, be more easier, convenient.

As the improvement of above embodiment, the number of multiple cameras is even number.Even number camera is evenly distributed on industry Producing line transports the two sides of straight line.At this point, it is not only in a center of symmetry between each camera, also at axial symmetry.Using even number camera Setting, further reduced the difficulty of assembly, the replacement of image acquisition subsystem.

It is shot simultaneously using multiple cameras, it can be in the image at synchronization acquisition glass difference edge.If adopted The number of camera is very few, then is not able to satisfy the requirement for obtaining the image at glass difference edge.And the camera used is excessive, this The mounting means of the relatively industrial producing line of a little units will become complicated, and and since number increases, corresponding cost can also increase Add.

Each camera has certain shooting angle β when shooting the glass to be measured of different location, only at β > 0 °, Camera can work normally.When using multiple cameras, shooting angle β > 0 ° of each camera when shooting is needed.Consider more The case where a camera is uniformly distributed, the number n and limit shooting angle β of camera_maxBetween relationship indicated by following formula:

β_max=90 ° -180 °/n

Only β_maxAt > 0 °, camera can be worked normally.Simultaneously as n must round numbers, by the formula it is found that n >= 3, i.e., at least need 3 cameras to be just able to satisfy shooting demand.At this point, at least two cameras, which are located at, transports the ipsilateral of straight line.

β_maxAlso it can regard the incident angle of camera, β as_maxIt is bigger, it is meant that the shooting direction of camera is more perpendicular to object table Face, shooting effect are better.

As can be seen that n is bigger from following table, β_maxBigger, average image quality is higher.That is, camera Number is more, and imaging effect is better

But more camera number, it is meant that higher cost and more complicated assembly difficulty.

According to many experiments as a result, the image quality of whole system is absolute when comprehensively considering the camera using different numbers Value, image quality improved values and hardware cost.The optimum state b of whole system is indicated with following formula:

B=8n-n²

By the formula it is found that as n=4, b obtains maximum value.

When n takes 4, camera be can be set on industrial producing line both sides, it is ensured that is in same level with glass to be measured.And And each camera since industrial producing line both sides are arranged between industrial producing line at a distance from it is identical, assembly can be reduced Difficulty.

Using 4 cameras, compared to 3 cameras, angle improves 50%.According to experimental result, image acquisition subsystem Camera from 3 to 4 when, the improvement of image quality is most.

The installation site of camera

As shown in Figure 6A, by taking 3 cameras as an example, Fig. 6 B-6D gives each camera in different image acquisition subsystems Relationship between blur-free imaging face.

As an implementation, the primary optical axis of three cameras is parallel to approximately the same plane.Such setting can prevent Inclination due to camera in longitudinal direction leads to the distortion of picture collected.

The blur-free imaging face P of each camera will meet: driving in industrial producing line and carry out relatively between target object and camera During movement, the every bit on target object surface can intersect with some blur-free imaging face P, that is: with vertical In the face of blur-free imaging face P be projection plane (transport plane in the present embodiment can be used as projection plane), blur-free imaging face P Projection approximation on a projection plane regards Projection Line Segment as, then the movement rail of any point on a projection plane on target surface All there are intersection points with the Projection Line Segment of blur-free imaging face P for mark.

Object is placed in industrial producing line, in terms of overlook direction, the side of the edge of object without departing from industrial producing line Edge.Such setting can guarantee no matter object is placed in industrial producing line with what direction, position, object Edge is all without departing from industrial producing line plane, that is to say, that: there is no the parts that can not be imaged completely on target surface (such as the part beyond industrial producing line plane cannot just be imaged completely).

In addition, due to no matter eye-observation or device imaging, can all exist and block before and after observed object itself Problem.Under the scene of shooting glass edge image, as shown in figures 6 b-6 c, when the direction that glass is put is in specific position, It will be blocked because of front and back, so that the complete image of glass edge to be measured can not be obtained.As shown in Figure 6B, under the scene, image Acquisition subsystem can not obtain to 39 right hand edge of rectangular glass image；Under the scene of Fig. 6 C, image acquisition subsystem without Method obtains the image of 39 left edge of rectangular glass to be measured.

According to test of many times, as shown in Figure 6 D, envelope is constituted between the line segment of multiple blur-free imaging face P on a projection plane In the case where closed region, as long as the detected product transported in industrial producing line is any one in the projection pattern on transport plane Point can be by the closed area, that is, product to be detected can completely pass through the closed area in transport process, that The complete clear image of the product to be detected can be obtained.

The multiple cameras for meeting above-listed condition can obtain the complete image of glass edge, and not have to consider that glass is placed Direction, position in industrial producing line.

The selection of best camera

The position of multiple cameras is fixed, that is to say, that the shooting angle of each camera is fixed.Since each edge of glass can Can there are problems that blocking, and glass is placed in industrial producing line with arbitrary angle, therefore, it is generally the case that be not Each camera can completely take certain or somewhere edge image of glass.

After knowing that subsystem obtains glass edge image and corresponding location information by position, to glass edge image into Row processing, obtains the normal vector at any one place on glass edge image.Position pointed by the normal vector, that is, it is directed to the point Best shooting angle.

If the glass has a plurality of edge, the average value of normal vector on each glass edge is calculated separately, and according to each The average value of the corresponding normal vector in edge determines optimal shooting angle to select optimal camera.

If the shape of the glass edge is a closed curve, can be by the normal direction of glass edge fixed intervals position Amount is ranked up according to the corner dimension of itself and certain reference direction, the optimal camera of the angle Selection of normal vector according to this.

As an alternative embodiment, the image of glass edge is a closed figure, the closed figure is obtained The normal vector of upper each point.The angular range of the corresponding normal vector of each camera can be identical or different, but all cameras pair The sum of angular range of normal vector answered, which is added, be not less than 360 °, to guarantee that any one place is all on any glass edge image There is at least one corresponding camera.The corresponding camera acquisition of normal vector when acquiring glass edge image, at the glass edge The image arrived is optimum image.

As a kind of optional embodiment, (point claims certain point of multiple cameras on industrial producing line width direction middle line Centered on point) be uniformly distributed, the angle that adjacent two lines between camera and central point are formed is consistent, and can also be described At multiple camera using central point to be uniformly distributed on the circumference of some circle in the center of circle.Such set-up mode, makes each Camera is all in a center of symmetry according to the central point.In the occasion comprising n camera, each camera selects in the range of 360 °/n Corresponding best normal vector.Such setting enables to the corresponding normal vector angular range of each camera identical, that is, The work load of each camera is identical, is conducive to the stable operation of image acquisition subsystem.As shown in Figure 3B, as a kind of reality Mode is applied, the industrial detection system based on Image Acquisition uses 4 image centers and is symmetrical arranged, and is evenly distributed in industrial producing line The two sides of straight line are transported, the angle between camera central axes is in 90 ° two-by-two.

Corresponding camera is selected according to the normal vector of glass edge, it is negative to reduce camera work on the work opportunity of control camera While load, the glass edge image of optimal imaging quality can be obtained.

The judgement of glass edge flaw

For glass edge to be measured, there are the glass edge images shot in the case of flaw as illustrated in figures 7 a-b.For difference The form of the glass of type, flaw can be different.

As an example, glass edge to be measured should cover frosted under normal circumstances in the present embodiment.At this point, glass The main Types of edge faults include " not covering frosted ", and flaw part generates mirror-reflection.Camera shoots this and does not cover frosted Flaw part obtain the bright, dark of image, angle and light source incidence direction depending on the corresponding mirror reflection surface.

If the reflective relation of light cannot be formed between source light, flaw part and camera, at this point, non-flaw part occurs A part of the light of light source is reflexed to camera by diffusing reflection；Mirror-reflection occurs for flaw part, since reflection angular direction is not right Quasi- camera, the light of light source do not pass through mirror-reflection and enter camera.As shown in Figure 7 A, in this case, the vision of flaw part Effect is darker than the visual effect of non-flaw part, so black is presented in the glass of the flaw part of non-frosted.

Fig. 7 B show glass edge to be measured in another case, and there are the images in the case of flaw.At this point, source light, Just geometry reflective relation is formed between flaw part and camera, the whole of source light is reflected into camera.Non- flaw part Still diffusing reflection occurs, a part of the light of light source is reflexed into camera.In this case, the visual effect of flaw part Visual effect instead than non-flaw part is brighter.Therefore, using such light source, the flaw portion of " not covering frosted " " highlighted " " dark " two kinds of situations may be had by being divided to.

As a kind of optional solution.When judging glass edge flaw, two different judgment thresholds are set, are wrapped Dark threshold value and bright threshold value are included, when the brightness value for judging glass edge somewhere is less than the dark threshold value, is then judged at this at for flaw； Or when the brightness value for judging glass edge somewhere is greater than the bright threshold value, also judge at this at for flaw.It can as another kind The solution of choosing.The average brightness value of whole glass edge is calculated first.Then by the brightness value of glass edge different location Subtract each other with the average brightness value, and absolute value is asked to the result subtracted each other.The absolute value is judged: if the big Mr. Yu of the absolute value One threshold value, then the brightness of the position is as excessive lightness or darkness, that is to say, that the position is mirror-reflection position, at this point, judgement should There is flaw in position.Brightness value mentioned here not only includes the brightness value in color image, also includes the ash in black white image Angle value.

Light source

When judging flaw using above-mentioned judgment mode, if the brightness at flaw was distinguished with the brightness at non-flaw Small, then the selection of threshold value is more difficult, and the judgement of flaw is also inaccurate.

In order to solve this problem, as an alternative embodiment, with reference to Fig. 8 A, on the blur-free imaging face P of camera Side's setting strip-shaped light source 81.In terms of depression angle, each strip-shaped light source can cover in image acquisition subsystem 32 one Or the blur-free imaging face P of multiple images acquisition unit.Strip-shaped light source may include the point light source being evenly distributed on strip-shaped light source Or area source.Using such setting, it can be ensured that when image acquisition subsystem acquires image, the clear part in image is all It can be irradiated by the source light of same brightness, angle, so that the luminance difference at each flaw and at non-flaw Absolute value increase, the brightness contrast at flaw and non-flaw is improved, to keep the judgement of flaw more accurate.Strip-shaped light source It, can be as best one can close to industrial producing line, to reduce the loss of brightness under the premise of object of which movement on not interfering industrial producing line.

The edge of ground glass and the intersection of glass surface are also possible to there are flaw, and this flaw is known as " chip ".

As another optional embodiment, the light that light source issues is directional light.Directional light covers the clear of camera Imaging surface P.Light source can be set beside camera, be in the ipsilateral of industrial producing line with camera.It can also set up separately with camera in work The two sides of industry producing line.It is arranged using such light source, the place that glass edge and glass surface cross can be irradiated to, so as to Glass " chip " enough is detected, improves the application range of on-line detecting system.

As another optional embodiment, with reference to Fig. 8 B, light source 82 is cyclic annular area source, the shape class of the area source It is similar to the side of cylindrical body or the side of cuboid.The light source surrounds glass to be measured, and the inner wall of entire light source all issues Even light.Such light source set-up mode can guarantee: light that light source inner wall issues, at glass edge flaw, Image Acquisition Mirror-reflection relationship can be formed between some camera three in system.At this point, no matter glass to be measured is in industrial producing line Put with what kind of direction, the flaw part of glass edge can by mirror-reflection by the light in somewhere on light source reflect into Some camera in image acquisition subsystem can be presented to make the visual effect of the image of glass edge flaw part It is highlighted.

As an alternative embodiment, light source form annular area source adds capping light source.This light-source structure phase Than can be improved the detection effect to " chip " in contain only annular light source the case where.

When light source is arranged, the lower edge at light source luminescent position to be at least consistent with the position of glass edge lower end or Lower than the lower end of glass edge, the effect of light source just can guarantee.But at the same time, such light source is possible to glass can be hindered to exist Movement in industrial producing line.

As an alternative embodiment, as shown in Figure 8 C, the side view of industrial producing line 30 at up-small and down-big trapezoidal shape, There is a highest plane in the process of running in industrial producing line 30.Annular light source or the annular light source of capping are around industrial producing line Highest plane setting.The position of 83 lower end of annular light source can flush in the face where the highest plane of industrial producing line 30 or Face where the highest plane of slightly below industrial producing line.Meanwhile the lower end of annular light source and the surface of industrial producing line 30 keep one Set a distance, the distance are greater than the maximum gauge of glass 39 to be measured.When glass 39 to be measured transports the highest plane of industrial producing line, Annular light source or the annular light source of capping surround glass to be measured.Due to the lower end of annular light source 83 and industrial producing line 30 keep away from From the maximum gauge for being greater than glass, can guarantee glass when being moved in industrial producing line without changing the position of annular light source, together When glass movement not will receive the interference of light source.It can also ensure that the lower edge at light source luminescent position is concordant or is lower than glass The lower end at edge, it is ensured that the integrality of light source incidence light angle.

As in fig. 8d, as another optional embodiment, annular light source 84 can move up and down.On annular light source End is provided with the telescopic rod 841 that can be moved up and down, which can pass through cylinder, hydraulic or motor driven.When industry produces When glass motion to be measured on line 30 is to 84 lower section of light source, the movement of industrial producing line stops, at this point, telescopic rod 841 declines, makes light The lower edge of source luminous site flushes in the lower end of glass 39 to be measured.After completing the acquisition of glass edge image, on telescopic rod 841 It rises, industrial producing line restores movement.Correspondingly, industrial producing line 30 can be driven by servo motor, it is arranged in industrial producing line and fixes Placement glass 39 to be measured region.Servo motor operates every time, and industrial producing line 30 is made to move fixed distance.It is such to set It sets, light source 84 can be made to be spaced up and down at the same time, to simplify program, reduce error probability.Meanwhile it can be with Press feedback device or distance-measuring device are set on light source, to ensure that light source will not damage glass 39 to be measured by pressure in decline.

Due to light source itself be it is opaque, in order to obtain the image of glass to be measured, image acquisition subsystem 32 is usually needed It is mounted between light source and glass to be measured.Simultaneously as the camera in image capturing system 32 is also impermeable under normal conditions Bright, camera often blocks some portion of light of light source.If camera includes multiple, and certain two camera and glass edge The relationship of reflection is just formd between flaw, then " dark " can be presented in the image of obtained glass edge flaw part again Situation.

To solve the above-mentioned problems, as a kind of optional solution, light source is set on camera 321.The light source can The side of camera lens is arranged in, and the form of area source can be set into the light source, so that the angle of emergent ray is wanted It at least can make up for it the light for the annular light source part that camera shelters from.The light source can also be annular, pacify around camera lens Dress.Such set-up mode, the light that can further ensure that light source issues can make up for it the light sheltered from by camera.

Defect Detection

After obtaining the clear image of glass edge, processor 33 detects flaw from image.

Defect Detection mode comprising Front camera

1, without the Defect Detection mode of image mosaic

Without the Defect Detection of image mosaic, clear image that each shooting time image acquisition subsystem is obtained into Row Defect Detection, to judge that it is indefectible that the image for taking pictures the glass edge somewhere obtained constantly has.Simultaneously as comprising preposition The detection system of camera 312 can directly obtain the position at any time glass edge where clear image, when detection is taken office Anticipate at the glass edge blur-free imaging at moment when there is flaw, also can know that the flaw place spatial position.Using this Mode can know that glass edge has the indefectible position with glass edge locating for the flaw simultaneously.

Specific deterministic process is as shown in Figure 9 A.

In step S911, a reference value of brightness or gray scale is determined.The determination of a reference value as an implementation can To be the average value of previous piece of glass edge brightness or gray scale.It as another embodiment, can before testing, by flawless The average value of brightness or gray scale that the glass of defect determines.

In step S912, the absolute value of the average value when previous picture brightness or gray scale and a reference value difference is calculated.

In step S913, judge whether the absolute value exceeds reasonable error range.If the absolute value of calculated result does not have Have beyond reasonable error range, then it is assumed that the position of glass edge does not occur flaw.If the absolute value of calculated result is super Reasonable error range is gone out, then it is assumed that flaw occurs in the position of glass edge, enters step S914.

In step S914, the spatial information of the corresponding glass edge of flaw picture is obtained.The spatial information, can be by position It sets and knows that subsystem 31 obtains.

In step S915, fed back.The mode of feedback can be alarmed by sound, light, can also pass through wired company The reversed work computer for being fed to quality inspection personnel, or it is sent to external mobile terminal by being wirelessly connected, such as smartwatch, intelligence On energy mobile phone.

It is indefectible directly to judge that the clear image of any time acquisition has, eliminates image mosaic step, efficiently quickly.

2, the Defect Detection mode after image mosaic

Without the Defect Detection mode of image mosaic, detect each time all relatively independent.Therefore, the result of detection is by preceding Face step is affected.For example a reference value depends on the brightness of the front glass of detection, grayscale information, in different ambient lightings Under the influence of can have difference, therefore with the variation of natural lighting, the precision of Defect Detection will receive influence.And any time Location information at blur-free imaging places one's entire reliance upon Front camera 312 and encoder 311, if Front camera 312 or encoder 311 information obtained itself produce deviation, the location information deviation that will lead to.

As another glass edge Defect Detection mode, as shown in Figure 9 B, steps are as follows:

In step S921, image is spliced, to form the clear image of complete glass edge.

In step S922, the brightness of the complete image obtained in step S921 or the average value of gray scale are calculated.When acquisition When picture is color image, which is average brightness；When the picture of acquisition is black and white picture, which is gray scale Average value.

Each picture in the complete image obtained in step S921 is calculated according to specific computation sequence in step S923 Element/pixel group brightness or gray scale, are compared with the average value of the brightness or gray scale that obtain in step S921.Manner of comparison can To be the absolute value for the brightness or gray scale and average value difference for calculating the pixel.It is also possible to calculate the brightness of the pixel or gray scale The absolute value being divided by with average value difference with average value, difference and average value are divided by, and can effectively exclude illumination condition to the flaw The judgement of defect position has an impact.If the brightness or gray scale and a reference value comparison result of comparing the pixel are greater than a certain threshold value, Think that the location of pixels is flaw location.

Flaw part is obtained in glass according to the sequence for calculating the flaw obtained in step S923 in step S924 The relative position at edge.At this point, also can receive position knows the spatial information that subsystem 31 obtains, it is opposite with what is obtained before It is integrated position.

In step S925, which feeds back quality inspection personnel.The mode of feedback can pass through wired connection The work computer of quality inspection personnel is fed back to, or is sent to external mobile terminal by being wirelessly connected, such as smartwatch, intelligence On mobile phone.

Defect Detection mode not comprising Front camera 312

1, without the Defect Detection mode of image mosaic

Without the Defect Detection of image mosaic, clear image that each shooting time image acquisition subsystem is obtained into Row Defect Detection, to judge that it is indefectible that the image for taking pictures the glass edge somewhere obtained constantly has.Due to not including preposition phase The Defect Detection mode of machine 312 is more difficult to get the position at any time blur-free imaging, and therefore, such detection mode is usually used In judging that it is indefectible that block glass edge to be measured has.

It is indefectible directly to judge that the clear image of any time acquisition has, eliminates image mosaic step, efficiently quickly.

2, the Defect Detection mode after image mosaic

Defect Detection is carried out after image mosaic, and detection accuracy on the one hand can be improved, it on the other hand, can be by obtaining the flaw Defect part compared to object under test other parts positional relationship, to obtain the location information of flaw part.Using this side Method can also obtain the position of flaw part although knowing the location information that system provides without position.

The basic principles and main features and advantage of the utility model have been shown and described above.The technical staff of the industry It should be appreciated that above-described embodiment does not limit the utility model in any form, it is all by the way of equivalent substitution or equivalent transformation Technical solution obtained, all falls in the protection scope of the utility model.

Claims (10)

1. a kind of foliated glass edge faults detection device based on Image Acquisition, comprising: roller frame, it is characterised in that:

The glass flaws detection device based on Image Acquisition further include: several transfer rollers, several detection cameras and place Manage device；

The transfer roller is rotatably connected to the roller frame to constitute one and be used to transmit foliated glass with one fixed width Transfer passages；

The pivot center of the transfer roller be arranged in parallel so that the foliated glass in the transfer passages can be supported on one it is defeated It transports in plane；

The detection camera be fixedly installed relative to the roller frame so that the camera lens for detecting camera is towards the transfer passages；

It is described detection camera be set to the transfer passages outside and at least there are two the detection camera be set to it is described defeated The heteropleural of wan access；

The optical axis of the detection camera lens lens is parallel to the transport plane and at least two is same positioned at the transfer passages The not parallel setting of optical axis of the camera lens of the detection camera of side；

The processor and the detection camera, which are constituted, to be electrically connected.

2. the foliated glass edge faults detection device according to claim 1 based on Image Acquisition, it is characterised in that:

The glass flaws detection device based on Image Acquisition further include: motor, the motor are fixedly attached to the roller frame, The electric machine shaft driving of the motor is connected to one or several to drive transfer roller to rotate in the transfer roller.

3. the foliated glass edge faults detection device according to claim 1 based on Image Acquisition, it is characterised in that:

The glass flaws detection device based on Image Acquisition further include: photoelectric sensor, the photoelectric sensor include: light Line emitting head and light receiver head, wherein light emitting head and light receiver head are separately positioned on opposite two of the transfer passages Side.

4. the foliated glass edge faults detection device according to claim 3 based on Image Acquisition, it is characterised in that:

Line between the light emitting head and light receiver head is parallel to the pivot center of the transfer roller.

5. the foliated glass edge faults detection device according to claim 4 based on Image Acquisition, it is characterised in that:

The updrift side for the transfer passages that the transfer roller is constituted is arranged in the light emitting head and light receiver head；The inspection Survey the downstream direction that the transfer passages that the transfer roller is constituted are arranged in camera.

6. the foliated glass edge faults detection device according to claim 1 based on Image Acquisition, it is characterised in that:

The foliated glass edge faults detection device based on Image Acquisition further include:

Front camera, is arranged in the top of the transport plane, and the camera lens of the Front camera is and described towards the transfer roller The primary optical axis of the camera lens of Front camera is vertical with the transport plane.

7. the foliated glass edge faults detection device according to claim 6 based on Image Acquisition, it is characterised in that:

The Front camera is line-scan digital camera, and the line-scan digital camera includes several photosensitive elements, the photosensitive element substantially edge It is parallel to the direction setting of the transfer roller.

8. the foliated glass edge faults detection device according to claim 7 based on Image Acquisition, it is characterised in that:

The glass flaws detection device based on Image Acquisition further include: photoelectric sensor, the photoelectric sensor include: light Line emitting head and light receiver head, wherein light emitting head and light receiver head are separately positioned on opposite two of the transfer passages Side；

The updrift side for the transfer passages that the transfer roller is constituted is arranged in the light emitting head and light receiver head；The inspection Survey the downstream direction that the transfer passages that the transfer roller is constituted are arranged in camera；The photoelectric sensor is arranged in the Front camera Between the detection camera.

9. the foliated glass edge faults detection device according to claim 1 based on Image Acquisition, it is characterised in that:

Lighting device is arranged in above the transmission plane that the transfer roller is constituted.

10. the foliated glass edge faults detection device according to claim 1 based on Image Acquisition, it is characterised in that:

The detection camera positioned at the transfer passages two sides that the transfer roller is constituted is symmetrical about the split of the transfer roller Setting.