CN101660894B - Device and method for multi-vision visual detection based on parallel light illumination - Google Patents

Abstract

The invention discloses a multi-eye vision detection device and method based on parallel light illumination, belonging to the technical field of intelligent machine vision. The measuring device includes a multi-degree-of-freedom adjustment and positioning mechanism and a parallel beam expanding optical mechanism. The multi-degree-of-freedom adjustment and positioning mechanism is mainly a gantry frame composed of a base, a threaded cylinder, a support nut and a guide rail beam, and three cameras and a parallel beam expanding light source are installed and fixed in it. The parallel beam expander optical mechanism is mainly composed of LED chip, lens group and mechanical structure, the purpose is to collimate and expand the divergent beam of LED point light source. The measurement method includes: positioning adjustment of the overall mechanism; position adjustment of the lens group in the parallel beam expander light source mechanism; light intensity adjustment of the parallel beam expander light source; calculation of matching parameters of camera images at three positions; and obtaining different differential images through image difference rules. The device of the invention has the characteristics of reliable performance, high degree of automation in the detection process, strong ability to resist common mode interference, high precision, etc., and provides a stable and reliable detection and measurement method for online production detection.

Description

Multi-eye visual inspection device and method based on parallel light illumination

technical field

The invention relates to a multi-eye vision detection device and method based on parallel light illumination, which are widely used in machine vision detection devices and methods in many fields such as machine manufacturing, on-line production, aerospace, and automobile body measurement.

Background technique

Intelligent machine vision technology is a new branch of automation developed in recent years, and it is an emerging detection technology. Based on machine vision, it integrates modern science and technology such as optoelectronics, computer technology, laser technology, and image processing technology to form a comprehensive detection system of light, machinery, electricity, and calculation. It has non-contact, full-field detection, high-speed The characteristics of high precision and high degree of automation can overcome the limitations of human capabilities, assist or replace humans in modern industrial production to complete online, automatic, continuous measurement and detection tasks, greatly improving the accuracy and real-time performance of the control system And reliability, so that product quality is effectively guaranteed.

In modern production, visual inspection is often an indispensable link. The development of machine vision technology can replace some tasks in artificial vision to a certain extent, especially in high-speed, continuous, mass production quality inspection, object recognition, defect extraction, Tasks that are difficult to complete manually such as contour acquisition and dimension measurement

At present, the common visual inspection methods are mostly developed on the basis of monocular vision technology. However, due to the complex and changeable environmental factors, a single image is often mixed with the influence of interference sources such as noise or stray light. It is difficult to accurately extract the optical information of the measured object. Taking the above factors into consideration, the present invention proposes a multi-eye vision detection device that can suppress environmental interference to a certain extent, eliminate common-mode noise of interference sources, and improve the stability and robustness of the detection system, and collect data from the multi-eye vision system The visual information of the camera is rich and multi-angle, so through the research and application of the algorithm based on multiple images, it can accurately calculate the surface or three-dimensional shape characteristics of the measured object, and through the calibration of the camera system, it is possible to obtain the image of the static object. related geometry.

Contents of the invention

The purpose of the present invention is to provide a multi-eye visual detection device and method based on parallel light illumination. The device of the present invention has the characteristics of reliable performance, high degree of automation in the detection process, strong anti-common-mode interference ability, and high precision. , physical three-dimensional shape measurement and aerospace field to provide stable and reliable detection and measurement methods.

A multi-eye visual detection device based on parallel light illumination provided by the present invention includes:

The multi-degree-of-freedom adjustment positioning mechanism is used for positioning the three cameras, and can arbitrarily adjust the three-direction degrees of freedom of the cameras in the plane.

The parallel beam expander optical mechanism is used to collimate and expand the divergent beam of the high-power LED point light source.

The multi-degree-of-freedom adjustment and positioning mechanism includes: a gantry frame composed of a base, threaded cylinders, support nuts and rail beams. The first threaded cylinder is fixed on the base and cooperates with the first support nuts. The first support nut carries the rail beam. The guide rail crossbeam is arranged with a first support slider, a second support slider and a third support slider that can slide and be fastened in position, each support slider has a structure composed of a second threaded cylinder and a second support nut . According to the order from left to right in Figure 1, the second support nut under the first slider on the left side supports a slot-shaped positioning adjustment housing, in which a parallel beam expanding white light source is fixed with screws, located in the vertical parallel expanding The direction of the beam of white light source Assemble the second trough-shaped adjustment and positioning shell on the bottom of the first trough-shaped adjustment and positioning shell, and place the first station camera in the vertical direction of the first light source through the positioning adjustment plate and positioning screws, and the indexing knob It is used to adjust the angle of the parallel beam expander white light source in the plane of the mechanism. The second supporting nut below the second slide block in the middle supports the groove-shaped positioning shell, and the second station camera is also fixed on the groove-shaped positioning casing, and ensures that the second station camera is vertically downward in the mechanism plane. The second support nut under the third slider on the right side also supports the third slot-shaped adjustment and positioning shell, and the third-station camera is placed in the support third slot-shaped adjustment and positioning shell, and the third-station camera can also be adjusted by the indexing knob Adjust its angle in the plane. A lifting adjustment table is placed on the marble base directly below the second station camera 12, and the workpiece to be measured is placed on the lifting adjustment table.

The parallel beam expander optical mechanism is composed of LED chip, lens group and mechanical structure, as shown in Fig. 2 . This mechanism ensures that the mechanical axis coincides with the optical axis, the center of the chip and the lens group is located on the optical axis, and the chip and the lens group are perpendicular to the optical axis. The high-brightness LED chip is placed on the rear side of the special LED condenser lens, the condenser lens reduces the divergence angle of the LED light source, and the condenser lens is fixed and locked in the first lens sleeve by a set screw. The lens L ₁ is clamped in the second lens sleeve, and the second sleeve can move axially in the first lens sleeve, and finally locked and fixed by a set screw. The third lens sleeve is used to clamp the aperture diaphragm, and can move along the optical axis in the second lens sleeve, and finally is positioned and locked by the positioning screw. The end of the third lens sleeve clamps the lens L ₂ and is fixed to the side of the square fixing frame of the half-mirror by screws. The semi-transparent and semi-reflective sheet is fixed along the diagonal direction of the side of the fixing frame.

A kind of multi-eye visual detection method based on parallel light illumination provided by the present invention includes the steps:

Overall organization positioning adjustment

According to the system design principle, in order to ensure the normal operation of the device, it is necessary to adjust the positions and angles of the three cameras in the system.

In the gantry structure, as shown in Figure 1, through the adjustment of the support nuts on the two studs, the guide rail beam can be adjusted to an appropriate height, and it can also move together with the lifting platform. Slide the 3 sliders on the beam, adjust their positions in the x direction, and try to keep the sliders at the same interval. The nuts under each slider can be adjusted to allow the load-bearing part below to move along the y direction. The camera at the first station can adjust its position up and down along the vertical direction of the light source in the trough-shaped adjustment and positioning housing, while the camera at the second station can adjust its position up and down in the vertical direction in the trough-shaped adjustment and positioning housing. The angle of the first station camera and the third station camera can be adjusted by the indexing knob. In order to obtain an ideal image effect, the optical axis direction of the light source and the optical axis direction of the third station camera should be adjusted to form a symmetrical angle with respect to the center normal of the measured workpiece.

Adjustment of Lens Group Position in Parallel Beam Expander Light Source Mechanism

According to the principle of parallel beam expansion, in order to obtain parallel light with good collimation, it is necessary to adjust the relative positions of the lens group and the LED chip. In Figure 2, after the chip LED passes through the front condenser lens, the divergence angle of the beam is smaller, and a virtual intersection point is obtained along the opposite direction of the beam, which is the point light source representing the LED chip. Adjust the lens sleeve with the set screw so that the distance from lens L ₁ to the virtual intersection point is twice the focal length of lens L ₁ . Adjust the lens sleeve by the set screw so that the distance from the aperture diaphragm to the lens _L1 is twice the focal length of the lens _L1 . The position of lens _L2 should be one focal length away from the aperture stop.

Light Intensity Adjustment of Parallel Expanded Beam Light Source

The high-brightness LED drive circuit chooses HV9910 as the main chip. This chip is based on the buck-boost constant current control principle, combined with the precise control method of PWM pulse width modulation, it can realize the continuous adjustment of the LED current in the range of 0-1A, and ensure the light intensity of the light source. continuous and stable change. During the operation of the system, adjust the input pulse duty cycle of the LED drive circuit, indirectly adjust the luminous intensity of the LED, so that the three-station camera can collect images in real time under the corresponding light intensity gear, and the obtained image group can be used for subsequent template extraction, image The difference algorithm provides the data.

Compute the matching parameters for the three location camera images

Due to the limitation of the positioning and adjustment accuracy of the mechanical positioning frame, before differential processing of multiple images, the matching parameters of each image need to be obtained first, including eccentric displacement parameters, plane rotation angle parameters and imaging scaling parameters. The present invention proposes a matching parameter determination method based on a two-dimensional Radon algorithm, and combines camera calibration technology to obtain camera external parameters for multiple image matching and differential processing.

Obtain different difference images through image difference rules

Through step 4, the matching parameters of the three images can be obtained, and the parameters can be used to adjust the image of the measured object so that its centroid coincides with the center of the image size, and rotate the long and short axes of the measured ellipse image to coincide with the x and y axes respectively. After the images are matched, the differential processing of the matched images is completed according to different image differential rules.

In order to obtain different matching and differential effects, the images obtained by the cameras of the first station and the second station under different brightnesses are combined and calculated according to certain criteria. The combination calculation criteria are designed as follows:

1.Pixels(Rule1)＝Pixels(Adding)-Pixels(750mA of Camera 2);

2.Pixels(Rule2)＝Pixels(Differential)-Pixels(750mA of Camera 2)*0.15;

3.Pixels(Rule3)＝Pixels(Adding)-Pixels(112mA of Camera 2)*0.58;

4.Pixels(Rule4)＝Pixels(Differential)*2.4-Pixels(112mA of Camera 2)*0.3;

5.Pixels(Rule5)＝Pixels(750mA of Camera 2)-Pixels(Differential);

6.Pixels(Rule6)＝Pixels(750mA of Camera 2)-Pixels(Adding)*0.5;

7.Pixels(Rule7)＝Pixels(112mA of Camera 2)-Pixels(Differential)*4.5;

8. Pixels(Rule8)＝Pixels(112mA of Camera 2)-Pixels(Adding)*1.35.

Compared with the prior art, the present invention has the following advantages:

High-power LED is used as the illumination source to irradiate the measured object with parallel light. At the same time, three cameras located at different positions are used to obtain images. The contour size information of the object is obtained through the matching and difference algorithm of multiple images to achieve two-dimensional plane geometry. purpose of measurement. Compared with monocular vision, this measurement technology obtains more complete image outline and defect information through difference algorithm, and the ability to suppress common mode interference is also more obvious, and it is easier to analyze geometric quantities. Compared with traditional monocular vision images The processing method is more adaptive and has higher precision.

Description of drawings

Fig. 1 is the overall mechanical structure diagram of the system of the present invention.

Fig. 2 is a structural diagram of the parallel collimated beam expander optical system of the present invention.

Fig. 3 is a working principle diagram of the system of the present invention.

Fig. 4 is a system control structure diagram of the present invention.

Fig. 5 is a system work flowchart of the present invention.

Fig. 6 is an effect diagram of matching difference in the present invention.

1-base, 2-threaded cylinder, 3-first support nut, 4-rail beam, 5-support slider (first, second and third), 6-second support nut, 7-threaded cylinder, 8 -Marble base, 9-lifting adjustment table, 10-workpiece under test, 11-first station camera, 12-second station camera, 13-third station camera, 14-parallel expanded beam white light source, 15- The first groove type adjustment positioning shell, 16-graduation knob, 17-the second groove type adjustment positioning shell, 18-positioning adjustment plate, 19-positioning screw, 20-the third groove type adjustment positioning shell.

21-high brightness LED, 22-LED front condenser lens, 23-positioning screw, 24-first lens sleeve, 25-lens L ₁ , 26-positioning screw, 27-second lens sleeve, 28-positioning Screw, 29-aperture stop, 30-the third lens sleeve, 31-lens L ₂ , 32-semi-transmissive and semi-reflective flat mirror, 33-square fixing frame for the semi-transmissive and semi-reflective mirror, 34-reflector.

Detailed ways

The technical solution of the present invention is further described in detail below in conjunction with accompanying drawings and specific examples:

A multi-eye visual detection device based on parallel light illumination provided by the present invention includes:

The multi-degree-of-freedom adjustment positioning mechanism is used for positioning the three cameras, and can arbitrarily adjust the three-direction degrees of freedom of the cameras in the plane.

The parallel beam expander optical mechanism is used to collimate and expand the divergent beam of the high-power LED point light source.

The multi-degree-of-freedom adjustment and positioning mechanism includes: a gantry frame composed of a base 1, a threaded cylinder 2, a support nut 3 and a rail beam 4, as shown in FIG. 1 . The threaded cylinder 2 is fixed on the base 1 and cooperates with the support nut 3 . The support nut 3 carries the rail beam 4 . The guide rail beam 4 is arranged with three support sliders 5 that can slide and be fastened in position, and each support slider 5 has a structure composed of a threaded cylinder 7 and a support nut 6 below. According to the order from left to right in Fig. 1, the nut 6 under the left slider 5 supports a groove-shaped positioning adjustment housing 15, and the parallel beam expanding white light source 14 is fixed with screws in the housing 15, which is located in the direction of the vertical light source 14 Assemble the groove-shaped adjustment and positioning shell 17 on the bottom of the shell 15, place the first station camera 11 in the vertical direction of the light source 14 through the positioning adjustment plate 18 and the positioning screw 19, and the indexing knob 16 is used to adjust the light source 14 on the mechanism plane inner angle. The nut 6 below the middle slider 5 supports the groove-shaped positioning housing 20, and the second station camera 12 is also fixed on the housing 20, and ensures that the camera 12 is vertically downward in the plane of the mechanism. The nut 6 under the right slide block 5 also supports the slot-type adjustment and positioning housing, and the third station camera 13 is placed in the housing, and the camera 13 can also adjust its angle in the plane by the indexing knob. A lifting adjustment table 9 is placed on the marble base 8 directly below the camera 12 at the second station, and the workpiece 10 to be measured is placed on the adjustment table 9 .

The parallel beam expander optical mechanism is composed of LED chip, lens group and mechanical structure, as shown in Fig. 2 . This mechanism ensures that the mechanical axis coincides with the optical axis, the center of the chip and the lens group is located on the optical axis, and the chip and the lens group are perpendicular to the optical axis. The high-brightness LED chip 21 is placed on the rear side of the special LED condenser lens 22 , the lens 22 reduces the divergence angle of the LED light source 21 , and the condenser lens 22 is fixed and locked in the lens sleeve 24 by a set screw 23 . The lens L ₁ 25 is clamped in the lens sleeve 27 , the sleeve 27 can move axially in the sleeve 24 , and is locked and fixed by the set screw 26 finally. The lens sleeve 30 is used to clamp the aperture stop 29 , and can move along the optical axis in the sleeve 27 , and is finally positioned and locked by the positioning screw 28 . The end of the sleeve 30 clamps the lens L ₂ 31 and is fixed to the side of the square fixing frame 33 of the half mirror by screws. The semi-transparent and semi-reflective sheet 32 is fixed along the diagonal direction of the side of the fixing frame 33 .

A kind of multi-eye visual detection method based on parallel light illumination provided by the present invention includes the steps:

1) Overall mechanism positioning adjustment

According to the system design principle, in order to ensure the normal operation of the device, it is necessary to adjust the positions and angles of the three cameras in the system. In the gantry structure, as shown in FIG. 1 , the guide rail crossbeam 4 can be adjusted to an appropriate height through the adjustment of the support nuts 3 on the two studs 2 , and it can also move together with the lifting platform 9 . Slide the three sliders 5 on the beam 4, adjust their positions in the x direction, and try to keep the sliders at the same interval. The nut 6 under each slide block can be adjusted so that the bearing part below it can move along the y direction. The first station camera 11 can adjust its position up and down along the vertical direction of the light source in the groove-shaped adjustment and positioning housing 17 , while the second station camera 12 can adjust its position up and down in the groove-shaped adjustment and positioning housing 20 along the vertical direction. The angles of the first station camera 11 and the third station camera 13 can be adjusted through the indexing knob 16 . In order to obtain an ideal image effect, the optical axis direction of the light source 14 and the optical axis direction of the third station camera 13 should be adjusted to form a symmetrical angle with respect to the center normal of the workpiece 10 to be measured.

2) Position adjustment of the lens group in the parallel beam expander light source mechanism

According to the principle of parallel beam expansion, in order to obtain parallel light with good collimation, it is necessary to adjust the relative positions of the lens group and the LED chip. In Fig. 2, after the chip LED 21 passes through the front condenser lens 22, the divergence angle of the light beam is smaller, and a virtual intersection point is obtained along the opposite direction of the light beam, which is the point light source representing the LED chip. The lens sleeve 27 is adjusted by the set screw 26 so that the distance from the lens L ₁ to the imaginary intersection is twice the focal length of the lens L ₁ . The lens sleeve 30 is adjusted by the set screw 28 so that the distance from the aperture stop 29 to the lens L ₁ is twice the focal length of the lens L ₁ . The position of the lens L ₂ should be one focal length away from the aperture stop 29 .

3) Light intensity adjustment of parallel beam expander light source

The high-brightness LED drive circuit chooses HV9910 as the main chip. This chip is based on the buck-boost constant current control principle, combined with the precise control method of PWM pulse width modulation, it can realize the continuous adjustment of the LED current in the range of 0-1A, and ensure the light intensity of the light source. continuous and stable change. During the operation of the system, adjust the input pulse duty cycle of the LED drive circuit, indirectly adjust the luminous intensity of the LED, so that the three-station camera can collect images in real time under the corresponding light intensity gear, and the obtained image group can be used for subsequent template extraction, image The difference algorithm provides the data.

4) Calculate the matching parameters of the three position camera images

Due to the limitation of the positioning and adjustment accuracy of the mechanical positioning frame, before differential processing of multiple images, the matching parameters of each image need to be obtained first, including eccentric displacement parameters, plane rotation angle parameters and imaging scaling parameters. The present invention proposes a matching parameter determination method based on a two-dimensional Radon algorithm, and combines camera calibration technology to obtain camera external parameters for multiple image matching and differential processing.

5) Obtain different differential images through image differential rules

Through step 4, the matching parameters of the three images can be obtained, and the parameters can be used to adjust the image of the measured object so that its centroid coincides with the center of the image size, and rotate the long and short axes of the measured ellipse image to coincide with the x and y axes respectively. After the images are matched, the differential processing of the matched images is completed according to different image differential rules.

The multi-eye visual detection device is composed of three cameras located in the same plane and with different acquisition directions and a parallel beam expanding light source mechanism, as shown in Figure 3.

In the plane of the mechanism, the parallel light source and camera 3 always keep the optical axis perpendicular, and the incident angle of the light beam is the angle between the optical axis of the light source and the horizontal line; camera 1 is located directly above the measured object to collect the image of the measured object vertically; the optical axis of camera 2 and The angle between the horizontal lines is the exit angle. Adjust the positions of the three cameras respectively so that the image of the measured object enters the camera field of view and is located near the center of the field of view, and ensure that the incident angle of camera 3 is equal to the exit angle of camera 2. Since the camera 3 maintains a vertical relationship with the light source, the light beam of the light source passes through the 45° semi-transparent and half-reflective mirror, and the light beam reflected by the measured object enters the light source and returns to the camera 3 . Adjust the distance of the camera along its own normal direction, so as to achieve the purpose of adjusting the imaging size and ensure that the image size in the field of view is similar, although the final image obtained by cameras 2 and 3 is the projection of the measured object onto the vertical plane of the optical axis, that is, the measured After projecting the intersecting line segment between the object surface and the vertical plane, the sinθ ratio will be shortened, but the size in the other vertical direction will not change. Therefore, the optical path of the camera can be adjusted through a rough estimation of the geometric quantity, and finally the size of the three images can be roughly adjusted to the same size. Then use the stable adjustment ability of the light source brightness to adjust the light source within a certain range of light intensity. When selecting the control current range, it should be ensured that camera 2 is not overexposed and camera 3 is not underexposed, and the aperture adjustment of cameras 2 and 3 can also be used in coordination to achieve a better exposure. After the camera is adjusted and fixed, under the different light intensity of the LED light source controlled by PWM, the synchronous acquisition of the images of the three cameras is completed through the acquisition system composed of the image acquisition card and the PC.

System control flow shown in Figure 4 . The PWM dimmer is controlled by outputting the duty ratio parameter through the software on the PC, so that the brightness of the light source can be continuously adjusted within a certain range. After the dimming system is stable each time, the image capture command is sent to the image acquisition card through the PC software, which controls the same time-domain image acquisition of the three-station camera.

The workflow of the system is shown in Figure 5. As shown in the flow chart, the PC obtains the image of the three-station camera through communication. Assuming that the measured object is a flat circle, as shown in Figure 3, the image obtained by camera 1 is a circle with a diameter of D*β, and β is the scaling factor of the camera; the images obtained by camera 2 and camera 3 are ellipses Shaped image, and the lengths of the major axis and the minor axis are D*β and D*β*sinθ respectively, and θ is the angle between the camera 2, 3 smooth and the horizontal line in the plane. Through the analysis of the profile features of the measured image and combined with the Radon algorithm, parameters such as the centroid position and rotation angle of the graphic can be calculated. The scaling factor can also be calculated by the ratio of the geometric quantity of the contour feature to the actual size, but considering that the smaller the θ angle is, the distance between the imaging planes of each point of the measured object is greater, so the image part of the measured object will be Clear, partially blurred. In order to obtain accurate calibration parameters of the camera, a target can be used, and the zoom ratio and depth of field of the camera at the position of the measured object can be calculated according to the precise graphics and size on the target.

According to the fact that different illumination brightness can be obtained by adjusting the LED drive current during the experiment, in this environment where the brightness can be adjusted, the image of the measured object in the cameras at three positions and angles is very different. Most of the beam of the light source obtained by the mirror reflection of the smooth plane of the measured object includes the light beam after removing the diffuse reflection of the non-smooth plane of the measured object, that is to say, the camera 2 shows the smooth surface of the measured object through high gray value The flat part, and the low gray value shows the defect plane or free-form surface part of the measured object. Here, the image obtained by camera 2 is called a bright field image. The position of camera 3 is symmetrical to the angle of camera 2, so the received image is in a complementary relationship with camera 2, that is, camera 3 shows the defect plane or free-form surface part of the measured object through high grayscale. Here, the image obtained by camera 3 is called Dark field image.

In order to obtain different matching difference effects, the images obtained by cameras 2 and 3 under different brightnesses are combined and calculated according to certain criteria. The combined calculation criteria are designed as follows:

1.Pixels(Rule1)＝Pixels(Adding)-Pixels(750mA of Camera 2);

2.Pixels(Rule2)＝Pixels(Differential)-Pixels(750mA of Camera 2)*0.15;

3.Pixels(Rule3)＝Pixels(Adding)-Pixels(112mA of Camera 2)*0.58;

4.Pixels(Rule4)＝Pixels(Differential)*2.4-Pixels(112mA of Camera 2)*0.3;

5.Pixels(Rule5)＝Pixels(750mA of Camera 2)-Pixels(Differential);

6.Pixels(Rule6)＝Pixels(750mA of Camera 2)-Pixels(Adding)*0.5;

7.Pixels(Rule7)＝Pixels(112mA of Camera 2)-Pixels(Differential)*4.5;

8.Pixels(Rule8)＝Pixels(112mA of Camera 2)-Pixels(Adding)*1.35;

Taking criterion 1 as an example, this criterion means adding the grayscale values of cameras 2 and 3 according to the appropriate matching coefficient and then subtracting it from the grayscale value of camera 2 at 750mA.

Application example

Taking a metal ring piece with a smooth surface with crack defects as an example, the images taken by the three-station cameras are obtained through the multi-eye visual inspection device, and the contour information of the measured object is obtained by the Radon algorithm, and then cut, translated, zoomed and matched and other methods to process the image to be differentiated. Finally, after processing according to the order of the above eight difference rules, the result in Figure 6 is obtained. It can be seen from the figure that the difference effects obtained by different criteria are completely different, and the contours are very clearly visible, which contains more information than the image obtained by monocular vision, and can eliminate the defect misjudgment caused by the uneven texture of the workpiece . In (1) and (2) in Figure 6, in addition to the prominent crack defects, the unevenness of the texture also interferes with the results. But in (3) and (4), this kind of interference is covered up, which shows that the frequency range of these two pictures is narrowed compared with the previous two pictures, so that the high-frequency information and low-frequency information are all compressed to the center of the frequency range, and the invalid Texture information interference. (5) and (7) The edge information is more obvious, the high-frequency information is prominent, and the low-frequency information is suppressed. Through the combined operation of these difference rules, effective image information that meets the requirements can be obtained. Compared with the monocular vision image processing algorithm, it is more adaptable, and the obtained contour and defect information are more obvious.

Claims (2)

1. used for multi-vision visual pick-up unit based on parallel light is characterized in that it comprises:

Multiple degrees of freedom is adjusted detent mechanism, is used to locate three cameras, and is used to regulate three direction degree of freedom of camera in the plane;

Parallel expanded beam optics mechanism is used for the divergent beams collimation of great power LED pointolite and expands bundle output;

Described multiple degrees of freedom is adjusted detent mechanism and is comprised: the gantry frame that base, thread cylinder, holding nut and guide rail beam are formed, and first thread cylinder is fixed on the base, and cooperates with first holding nut; The first holding nut bearing guide rail crossbeam, guide rail beam has can slide and locate the first fastening support slipper, second support slipper and the 3rd support slipper, below each support slipper by the structure of forming by second thread cylinder and second holding nut, second holding nut below first support slipper of left side supports first grooved location and adjusts shell, first grooved location is adjusted in the shell and has been screwed parallel expansion bundle white light source, be positioned at vertical parallel and expand the bottom that the direction of restrainting white light source is assemblied in second grooved adjustment location shell on first grooved adjustment location shell, the first station camera is placed to the vertical direction of first light source by location adjustment flat board and dog screw, the calibration knob is used for regulating parallel expansion bundle white light source angle in the mechanism plane, second holding nut below middle second support slipper supports second grooved adjustment location shell, the second station camera also is fixed on the shell of second grooved adjustment location, and guarantee the second station camera in the mechanism plane vertically down, second holding nut under the 3rd support slipper of right side also supports the 3rd grooved adjustment location shell, lay the 3rd station camera in supporting the 3rd grooved adjustment location shell, the 3rd station camera is regulated its angle planar by the calibration knob; Be positioned on the marble base under the second station camera (12) and lay a lift adjustment platform, measured workpiece is positioned on the lift adjustment platform;

Described parallel expanded beam optics mechanism is made up of led chip, lens combination and physical construction, and this mechanism guarantees that mechanical axle center overlaps with optical axis, and described chip and lens combination are centered close on the optical axis, and described chip and lens combination vertical optical axis; Described chip is positioned over special-purpose LED collector lens rear side, and collector lens diminishes the angle of divergence of led light source, collector lens in first lens sleeve by the dog screw fixing and locking, lens L ₁Be held in second lens sleeve, second sleeve can move in first lens sleeve vertically, is locked by dog screw at last; The 3rd lens sleeve is used for the clamping aperture diaphragm, and moves along optical axis direction in second lens sleeve, at last by the dog screw positioning and locking; The end of the 3rd lens sleeve clamps lens L ₂, and by the side of screw retention in the square fixed mount of semi-transparent semi-reflecting lens, semi-transparent semi-reflecting eyeglass is fixed along fixed mount side diagonal.

2. used for multi-vision visual detection method based on parallel light is characterized in that the step that comprises:

1) whole mechanism location adjustment, the position and the angle of three cameras in the regulating system

In the framework of gantry, the adjusting by the holding nut on two double-screw bolts is adjusted to suitable height with guide rail beam, and cooperate lifting table to move together, slip is positioned at three slide blocks on the crossbeam, adjusts their positions in the x direction, keeps slide block identical at interval as far as possible; Regulate the nut under each slide block, make below the slide block bearing part along y to moving, the first station camera in the shell of first grooved adjustment location along the vertical direction up-down adjustment position of light source, and second station camera up-down adjustment position vertically in the shell of second grooved adjustment location, the first station camera and the 3rd station camera be by calibration knob adjusting angle, and the optical axis direction of regulating light source and the optical axis direction of the 3rd station camera become symmetry angle about the centre normal of measured workpiece;

2) the lens combination position is adjusted in the parallel expansion light source beam mechanism

Adjust the relative position of lens combination and led chip, make chip LED pass through preposition collector lens after, beam divergence angle is littler, obtains an imaginary intersection point along the opposite direction of this light beam, this point is the pointolite of representing led chip; Regulate lens sleeve by dog screw and make lens L ₁Distance to imaginary intersection point is lens L ₁The two focus length place regulates lens sleeve by dog screw and makes aperture diaphragm to lens L ₁Distance be lens L ₁The two focus length place, lens L ₂The position should be the one times of focal length place of self apart from aperture diaphragm;

3) parallel expansion light source beam light intensity regulating

The high-brightness LED driving circuit selects HV9910 as master chip, in conjunction with PWM width modulation accuracy control method, realizes the continuous adjusting of LED electric current in 0～1A scope guaranteed the variation of the continous-stable of light source intensity; Regulate the input pulse dutycycle of led drive circuit, the luminous intensity of indirect regulation LED makes three station cameras real-time images acquired under corresponding light prime time position, and the image sets that obtains provides data for follow-up template extraction, image difference algorithm;

4) matching parameter of three position camera images of calculating

Before to the multiple image difference processing, at first obtain the matching parameter of each width of cloth image, comprise center displacement parameter, plane corner parameter and imaging scaling parameter, use the matching parameter decision method of two-dimentional Radon algorithm, and combining camera calibration technique acquisition camera external parameter, so that multiple image coupling difference processing;

5) obtain different difference images by the image difference rule

Can obtain the matching parameter of three width of cloth images by step 4, parameter is just adjusted to the measured object image its centre of form and is overlapped with the picture size center thus, and rotate tested oval image major and minor axis and overlap with x, y axle respectively, with images match well after, again according to different image difference rules finish to mate the back image difference processing; The camera (3) of first station under the different brightness and the camera (2) of the 3rd station are obtained image according to the combination of criteria computing, and the combination of criteria computing is to handle by following eight difference rule orderings:

(1)Pixels(Rule1)＝Pixels(Adding)-Pixels(750mA?of?Camera?2)；

(2)Pixels(Rule2)＝Pixels(Differential)-Pixels(750mAofCamera2)*0.15；

(3)Pixels(Rule3)＝Pixels(Adding)-Pixels(112mA?of?Camera?2)*0.58；

(4)Pixels(Rule4)＝Pixels(Differential)*2.4-Pixels(112mA?of?Camera?2)*0.3；

(5)Pixels(Rule5)＝Pixels(750mA?of?Camera?2)-Pixels(Differential)；

(6)Pixels(Rule6)＝Pixels(750mA?of?Camera?2)-Pixels(Adding)*0.5；

(7)Pixels(Rule7)＝Pixels(112mA?of?Camera?2)-Pixels(Differential)*4.5；

(8)Pixels(Rule8)＝Pixels(112mA?of?Camera?2)-Pixels(Adding)*1.35；

Wherein, the camera (3) of Pixels (Adding) expression first station and camera (2) the gradation of image value of the 3rd station are according to the matching factor addition, and Pixels (750mA of Camera 2) represents the gray-scale value of the camera (2) of the 3rd station.

Images