CN106018422A - Matching-based visual outline defect inspection system and method for specially-shaped stamping parts - Google Patents

Abstract

The invention provides a matching-based visual outline defect inspection system and method for specially-shaped stamping parts. The matching-based visual outline defect inspection system is characterized by comprising an intelligent camera, a first support, a second support, a computer, a control part, a transparent rotary disc, a motor, a sensor and a back light source, wherein the control part controls rotation of the motor and triggering of the intelligent camera; the intelligent camera, the first support and the computer together constitute a visual system; the intelligent camera is fixed above the transparent rotary disc through the first support and connected with the computer and the control part through communication lines; the computer is connected with the control part; the transparent rotary disc is connected with the motor; the back light source is located right below the intelligent camera and also below the transparent rotary disc; the sensor comprises a sensor transmitter and a sensor receiver which are oppositely arranged and located on the same horizontal line, and the sensor transmitter is connected with the second support.

Description

Special-shaped stamping parts profile defects vision detection system based on coupling and method

Technical field

The present invention relates to determine the matching detection systems technology field of workpiece profile area and profile not bending moment, specifically one Plant special-shaped stamping parts profile defects vision detection system based on coupling and method.

Background technology

Along with the large-scale production of product, a lot of stamping parts have the features such as batch is big, parts are little, specification is many, to product Quality testing required precision more and more higher, in order to ensure the quality of automobile workpiece, the defects detection to workpiece is essential Link.Stamping parts was carried out visually or kind of calliper by the method frequently with artificial cognition in the past, but this detection method speed Slowly, efficiency is low, labor strength is big, and quality is also difficult to ensure that.Therefore, computer picture detection technique is applied to stamping parts The detection of quality has important demand and value.Machine vision have noncontact, high efficiency, in high precision, easy of integration etc. notable Advantage, application machine vision carries out the defects detection of workpiece, can solve to perplex a lot of problems of enterprise, it is ensured that detection Seriality and real-time.Currently, a lot of detection projects conform to the principle of simplicity singly to move towards complicated, and detection mode is also from manually moving towards automatization very To intelligent.Vision-based detection is exactly a kind of trend of detection, and machine vision solution can utilize machine to replace human eye to do Various measurements and judgement, have noncontact, adaptable, rapidly and efficiently, accurate, flexible, reliability high, existing For industrial detection receives extensively attention.

Chinese patent CN103914827 discloses the visible detection method of a kind of weather strip for automobile profile defects, the method By loading image is carried out Threshold segmentation, obtain bianry image, carry out rim detection, extract the seat of each edge pixel point Mark, calculates the convex closure of target area, calculates the minimum enveloping surface product moment shape in sealing strip cross section, by area, and edge pixel Point judges whether defect, and it is disadvantageous in that: detection method based on contour area needs sample several are determined in advance What shape is regular, and the pixel shared by the same workpiece image that diverse location collects under camera fields of view can change Become, and owing to the angle difference that defect position is different, sample is put all can produce impact to the estimation of preferable area, make Become testing result inaccurate, it is impossible to reach the accuracy requirement judged.

Summary of the invention

For the deficiencies in the prior art, the technical problem that the present invention intends to solve is to provide a kind of abnormity punching based on coupling Casting die profile defects vision detection system and method.This detecting system is (the abnormity stamping parts letter of 4mm abnormity stamping parts according to thickness Claim special-shaped workpiece or workpiece) design, the single-frame images collected according to smart camera, extracts interesting target abnormity stamping parts Profile, by the Hu square of effective contour curve calculating profile, (Hu square has rotation, translation invariance is also called Hu not bending moment, i.e. For not bending moment) and the area of interior zone.In gatherer process, owing to workpiece exists certain thickness, different at camera fields of view Position, the profile of the identical workpiece collected there are differences, be not to fit like a glove.For this problem, take multiple sample The method of this coupling, calls multiple sample form Retention area feature, Hu square parameter.Bonded area difference matching detection and Hu Invariant moments matching detection algorithm, and then complete real-time, continuous profile defects detection and classification.The method, it is possible to accurately detect Go out the defect of account for whole workpiece area more than 0.5%, improve accuracy of detection, it is possible to meet abnormity stamped workpieces defects detection Requirement.

The technical scheme is that,

A kind of special-shaped stamping parts profile defects vision detection system based on coupling, it is characterised in that this detecting system includes Smart camera, the first support, the second support, computer, control part, transparent rotating circular disk, motor, sensor and back light Source, controls part and controls rotation and the triggering of smart camera of motor；Collectively formed by smart camera, the first support and computer Visual system, described smart camera is fixed on the top of transparent rotating circular disk by the first support, and smart camera passes through order wire Being connected with computer and control part, computer is connected with controlling part, and described transparent rotating circular disk is connected with motor, institute simultaneously State backlight and be positioned at the underface of smart camera, and be in the lower section of transparent rotating circular disk；Described sensor includes sensor Emitter and transducer receivers, transducer receivers is oppositely arranged with sensor transmitter, and in the same horizontal line, passes Sensor emitter and the second support are connected, and transducer receivers is fixed on the first support bottom, transducer receivers and sensor Emitter is connected with control part simultaneously.

A kind of special-shaped stamping parts profile defects visible detection method based on coupling, uses above-mentioned detecting system, the party Method utilizes the standard extracted and the profile information of abnormity stamping parts to be detected, exists according to utilizing standard workpiece and defective workpiece Area discrepancy and there is rotations, translate, scale the difference of the profile Hu not bending moment of invariance, contrast special-shaped punching press to be detected Part, the area of standard special-shaped workpiece and profile not bending moment, be calculated the matching rate (difference in areas of difference in areas and profile respectively It is the result of area difference algorithm), then set threshold value respectively, it is judged that abnormity stamping parts whether existing defects, the tool of the method Body step is:

The first step, image procossing

1-1 Image Acquisition: obtain contrast significantly abnormity stamping parts image by smart camera；

1-2 image threshold adaptivenon-uniform sampling: on the basis of step 1-1, calculates each pixel for the image collected The weighted mean in point 5 × 5 regions around, deducting a constant obtains adaptive threshold, and each pixel pixel value is more than threshold Value, it is defined as the available point of interesting target object abnormity stamping parts in image；

Seeking of 1-3 profile takes: on the basis of step 1-2, has interesting target object abnormity stamping parts in image Effect point carries out the continuous evolution of curve, arranges iterations, can obtain effective contour curve C (x, y) figure of abnormity stamping parts Picture；

Second step, image information analysis

The calculating of 2-1 area: the contour curve C that obtains according to step 1-3 (x, y) is divided into two parts by image-region, A part is that (x, y), another part is that (x, y), to obtain for contour curve perimeter outC to contour curve interior zone inC (x, y) is integrated contour curve interior zone inC, i.e. can obtain characterizing the area information of contoured interior area size, this step Suddenly the area S of standard abnormity stamping parts can be calculated₀Area S with special-shaped stamping parts to be detected_n, wherein, S_nRefer to The area of n workpiece to be detected, n >=1；

The calculating of 2-2 profile Hu not bending moment: (x, y), by owning on profile for the contour curve C obtained for step 1-3 Point carries out integral operation, obtains profile Hu not bending moment, it is assumed that the some number on profile is N, obtains profile p+q rank by formula (3) Central moment,

${\mathrm{\μ}}_{pq}=\underset{i=1}{\overset{N}{\Σ}}{(x-\stackrel{\‾}{x})}^p{(y-\stackrel{\‾}{y})}^{q}f(x,y);---\left(3\right)$

In formula, the square on p correspondence x dimension, the square on q correspondence y-dimension,WithRepresent the center of gravity of profile:

Normalized p+q central moment η_pqIt is defined as:

${\mathrm{\η}}_{pq}={\mathrm{\μ}}_{pq}/\left({\mathrm{\μ}}_{00}^{\mathrm{\ρ}}\right);---\left(6\right)$

Wherein, ρ=(p+q)/2+1；

Hu not bending moment is the linear combination of the normalization central moment obtained by formula (6), can obtain profile by formula (3) Second order and third central moment, be brought into formula (6) and obtain second order three rank normalization central moment, by second order and three normalization centers, rank Square seven Hu not bending moment I of structure₁～I₇, the formula of concrete Hu not bending moment is formula (7), and Hu not bending moment can under the conditions of consecutive image Keep translation, scaling and invariable rotary；:

$\begin{array}{c}{I}_1={\mathrm{\η}}_{20}+{\mathrm{\η}}_{02};\\ I_2={({\mathrm{\η}}_{20}-{\mathrm{\η}}_{02})}^2+4{\mathrm{\η}}_{11}^2;\\ I_3={({\mathrm{\η}}_{30}-{\mathrm{\η}}_{12})}^2+{(3{\mathrm{\η}}_{21}-{\mathrm{\η}}_{03})}^2;\\ I_4={({\mathrm{\η}}_{30}+{\mathrm{\η}}_{12})}^2+{({\mathrm{\η}}_{21}+{\mathrm{\η}}_{03})}^2;\\ I_5=({\mathrm{\η}}_{30}-3{\mathrm{\η}}_{12})({\mathrm{\η}}_{30}+{\mathrm{\η}}_{12})({\left({\mathrm{\η}}_{30}+{\mathrm{\η}}_{12}\right)}^2-3{\left({\mathrm{\η}}_{21}+{\mathrm{\η}}_{03}\right)}^2)\\ +(3{\mathrm{\η}}_{21}-{\mathrm{\η}}_{03})({\mathrm{\η}}_{21}+{\mathrm{\η}}_{03})(3{\left({\mathrm{\η}}_{30}+{\mathrm{\η}}_{12}\right)}^2-{\left({\mathrm{\η}}_{03}+{\mathrm{\η}}_{12}\right)}^2);\\ I_6=({\mathrm{\η}}_{20}-{\mathrm{\η}}_{02}){({\mathrm{\η}}_{30}+{\mathrm{\η}}_{12})}^2-{({\mathrm{\η}}_{03}+{\mathrm{\η}}_{21})}^2+4{\mathrm{\η}}_{11}({\mathrm{\η}}_{30}+{\mathrm{\η}}_{12})({\mathrm{\η}}_{03}+{\mathrm{\η}}_{12});\\ I_7=(3{\mathrm{\η}}_{21}+{\mathrm{\η}}_{03})({\mathrm{\η}}_{30}+{\mathrm{\η}}_{12}){({\mathrm{\η}}_{12}+{\mathrm{\η}}_{30})}^2-3{({\mathrm{\η}}_{21}+{\mathrm{\η}}_{03})}^2\\ +(3{\mathrm{\η}}_{12}-{\mathrm{\η}}_{30})({\mathrm{\η}}_{21}+{\mathrm{\η}}_{03})(3{\left({\mathrm{\η}}_{30}+{\mathrm{\η}}_{12}\right)}^2-{\left({\mathrm{\η}}_{21}+{\mathrm{\η}}_{03}\right)}^2);\end{array}---\left(7\right)$

2-3 area Difference Calculation: the area S of the standard abnormity stamping parts that step 2-1 is obtained₀With abnormity to be detected punching press The area S of part_nDiffer from, obtain the area difference result Δ S of abnormity stamping parts to be detected and standard abnormity stamping parts_n, i.e. Δ S_n= |S₀-S_n|；

The calculating of 2-4 outline rate: the Hu not bending moment I obtained according to step 2-2_i(1≤i≤7), mark is passed judgment in definition Accurate:

$I(A,B)=\underset{i=1}{\overset{7}{\Σ}}|\frac{1}{{m_i}^A}-\frac{1}{{m_i}^B}|;---\left(9\right)$

Wherein, m_i ^A、m_i ^BIt is defined as:

m_i ^A=sign (I_i ^A)·log|I_i ^A|； (10)

m_i ^B=sign (I_i ^B)·log|I_i ^B|； (11)

Ask for the matching rate (in units of percentage ratio) of standard workpiece and workpiece to be detected:

I=100-I (A, B) × 100； (12)

In formula, A, B refer to standard workpiece and workpiece to be checked respectively；

3rd step, detects in real time

3-1 image real-time acquisition: after detecting system starts, controls part and controls the transparent rotating circular disk of carrying workpiece with perseverance Constant speed degree rotates counterclockwise, when workpiece motion s is to when having between the correlation position of sensor transmitter and transducer receivers, passes Collection signal is transferred to control part by sensor, and then the transparent rotating circular disk controlling carrying workpiece stops 1s, it is thus possible to work Part carries out stationary picture；Control part simultaneously is signaled to smart camera, utilizes smart camera external trigger pattern to trigger intelligence Camera Real-time Collection workpiece image, calls template image to carry out next step image procossing；

3-2 image procossing: after image acquisition terminates, carries out image procossing by computer, by the profile of step 2-4 Joining rate judgment criteria, the area difference result in conjunction with step 2-3 sets threshold value, carries out matching detection, it is judged that whether workpiece closes Lattice, and feed back to control part；

3-3 workpiece is classified: after image procossing has judged that workpiece is the most qualified, next by outside grasping mechanism to work Part captures, and qualified storehouse put into by qualified workpiece, otherwise puts into recovery storehouse, thus completes the detection of whole workpiece, defect.

Compared with prior art, the invention has the beneficial effects as follows:

The inventive method is by controlling the transparent rotating circular disk stop and start of part Based Intelligent Control carrying workpiece, it is possible to protect Card smart camera can carry out Real-time Collection to it under workpiece resting state, it is to avoid produced workpiece is captured in motion continuously Anamorphose processes, to successive image, this problem that exerts an adverse impact.The image that acquisition for contrast is obvious, and pass through Adaptive threshold fuzziness can accurately obtain the available point of interesting image target special-shaped workpiece, takes offer for accurately seeking of profile Precondition.The active profile selectivity segmentation that the present invention uses obtains the method for profile and does not relies on gradient judgement, can be very well Improve rim detection extract each pixel time exist edge leakage problem, for border rough or discontinuous also It is capable of detecting when have the strongest anti-noise, capacity of resisting disturbance, strong robustness.Use multiple sample to mate, and it is constant to combine Hu The rotation of square, translate, scale invariance, solve area difference and mate inaccurate problem, significantly improve the real-time of system Property and adaptability.At real-time detection-phase, it is possible to the most effectively workpiece signal to be sent to control system, Based Intelligent Control image Gather, and then complete the defects detection of abnormity stamping parts, it is not necessary to manual intervention, improve work efficiency.

The use field of the present invention and significance be:

This method is applicable to the profile defects detection of abnormity stamping parts, and abnormity stamping parts occupies act foot in industrial sector chain The status of weight, is widely used in electronic device, automobile, main equipment, ornament materials etc., and the accurate Intelligent Measurement of method of needing badly is rushed The profile defects of casting die, and improve the accuracy of detection.Therein it is crucial that the accurate of profile obtains, and matching algorithm Accuracy.Vision detection system of the present invention, it can be avoided that dependence to gradient, obtains and has high anti-noise, anti-interference, strong robustness Contour curve, solve profile obtain exist accuracy, rapidity problem.Contour area difference is mated by the inventive method Detection and Hu invariant moments matching detection algorithm combine, and substantially increase accuracy and the adaptability of detection, it is possible to meet and sentence The requirement of disconnected accuracy, is more suitable for commercial Application.

Accompanying drawing explanation

Fig. 1 is that the structure of the present invention special-shaped stamping parts a kind of embodiment of profile defects vision detection system based on coupling is shown It is intended to；

Fig. 2 is that the present invention special-shaped stamping parts profile defects vision detection system one embodiment based on coupling regards at camera Under open country, gather standard stamping parts template schematic diagram when being in diverse location；

Fig. 3 is the structural representation of a kind of abnormity stamping parts of detection of the present invention；

Fig. 4 is that the present invention special-shaped stamping parts profile defects vision detection system one embodiment workpiece based on coupling is to passing After sensor location triggered camera, stamping parts fixed range schematic diagram under the visual field；

Fig. 5 is that the present invention special-shaped stamping parts profile defects vision detection system one embodiment based on coupling adds noise The contour images schematic diagram of rear extraction；

In figure, 1 first support, 2 second supports, 3 smart cameras, 4 backlight, 5 sensor transmitter, 6 sensors connect Receive device, 7 motors, 8 transparent rotating circular disks, 9 direction of rotation, 10 abnormity stamping parts, 11 order wires, 12 computers.

Detailed description of the invention

Below in conjunction with embodiment and accompanying drawing, the invention will be further described.

The present invention special-shaped stamping parts profile defects vision detection system based on coupling (is called for short detecting system or system, ginseng See Fig. 1) include smart camera the 3, first support the 1, second support 2, computer 12, control part, transparent rotating circular disk 8, motor 7, sensor and backlight 4, controls part and controls rotation and the triggering of smart camera 3 of motor 7；By smart camera 3, One support 1 and computer 12 collectively form visual system, and described smart camera 3 is fixed on transparent rotational circle by the first support 1 The top of dish 8, smart camera 3 is connected with computer 12 and control part by order wire 11 simultaneously, computer 12 and control portion Point connect, for show smart camera 3 gather and process after image；Described transparent rotating circular disk 8 is connected with motor 7, described Backlight 4 is positioned at the underface of smart camera 3, and is in the lower section of transparent rotating circular disk 8；Described sensor includes sensor Emitter 5 and transducer receivers 6, transducer receivers 6 is oppositely arranged with sensor transmitter 5, and at same level line On, sensor transmitter 5 is connected with the second support 2, fixes position and completes the accurate transmission of signal, and transducer receivers 6 is solid Being scheduled on the first support 1 bottom, transducer receivers 6 is connected with control part with sensor transmitter 5 simultaneously, and the two has coordinated The collection of stamping parts position signalling.

It is 3～5mm that present system is further characterized by the thickness of described abnormity stamping parts 10, and stamping parts is a length of 40～60mm, width is 15～55mm, and it is more than 0.5% that the rejected region that can detect that accounts for the minimum percent of entirety.

The present invention special-shaped stamping parts profile defects visible detection method (abbreviation method) based on coupling uses above-mentioned inspection Examining system, is mainly used in industry spot, the detection of the profile of abnormity stamping parts, identifies and judges profile defects and then judge punching Casting die is carried out with or without defect situation simultaneously, and detection algorithm fully adapts to field condition, and the method utilizes the standard extracted and to be checked Survey the profile information of abnormity stamping parts, according to the area discrepancy utilizing standard workpiece and defective workpiece to exist and have rotation, Translation, the difference of profile not bending moment of scaling invariance, contrast abnormity stamping parts to be detected, the area of standard special-shaped workpiece and Profile not bending moment, is calculated the matching rate of difference in areas and profile respectively, then sets threshold value respectively, it is judged that abnormity stamping parts is No existing defects, comprising the concrete steps that of the method:

The first step, image procossing

1-1 Image Acquisition: obtain contrast significantly abnormity stamping parts image by smart camera 3；

1-2 image threshold adaptivenon-uniform sampling: on the basis of step 1-1, calculates each pixel for the image collected The weighted mean in point 5 × 5 regions around, deducting a constant obtains adaptive threshold, and each pixel pixel value is more than threshold Value, it is defined as the available point of interesting target object abnormity stamping parts in image；

Seeking of 1-3 profile takes: on the basis of step 1-2, has interesting target object abnormity stamping parts in image Effect point carries out the continuous evolution of curve, arranges iterations, can obtain effective contour curve C (x, y) figure of abnormity stamping parts Picture；

Second step, image information analysis

The calculating of 2-1 area: the contour curve C that obtains according to step 1-3 (x, y) is divided into two parts by image-region, A part is that (x, y), another part is that (x, y), to obtain for contour curve perimeter outC to contour curve interior zone inC (x, y) is integrated contour curve interior zone inC, i.e. can obtain characterizing the area information of contoured interior area size, this step Suddenly the area S of standard abnormity stamping parts can be calculated₀Area S with special-shaped stamping parts to be detected_n, wherein, S_nRefer to The area of n workpiece to be detected, n >=1；

The calculating of 2-2 profile Hu not bending moment: (x, y), by owning on profile for the contour curve C obtained for step 1-3 Point carries out integral operation, obtains profile Hu not bending moment, it is assumed that the some number on profile is N, obtains profile p+q rank by formula (3) Central moment,

${\mathrm{\μ}}_{pq}=\underset{i=1}{\overset{N}{\Σ}}{(x-\stackrel{\‾}{x})}^p{(y-\stackrel{\‾}{y})}^{q}f(x,y);---\left(3\right)$

In formula, the square on p correspondence x dimension, the square on q correspondence y-dimension,WithRepresent the center of gravity of profile:

Normalized p+q central moment η_pqIt is defined as:

${\mathrm{\η}}_{pq}={\mathrm{\μ}}_{pq}/\left({\mathrm{\μ}}_{00}^{\mathrm{\ρ}}\right);---\left(6\right)$

Wherein, ρ=(p+q)/2+1；

Hu not bending moment is the linear combination of the normalization central moment obtained by formula (6), can obtain profile by formula (3) Second order and third central moment, be brought into formula (6) and obtain second order three rank normalization central moment, by second order and three normalization centers, rank Square seven Hu not bending moment I of structure₁～I₇, the formula of concrete Hu not bending moment is formula (7), and Hu not bending moment can under the conditions of consecutive image Keep translation, scaling and invariable rotary；:

$\begin{array}{c}{I}_1={\mathrm{\η}}_{20}+{\mathrm{\η}}_{02};\\ I_2={({\mathrm{\η}}_{20}-{\mathrm{\η}}_{02})}^2+4{\mathrm{\η}}_{11}^2;\\ I_3={({\mathrm{\η}}_{30}-{\mathrm{\η}}_{12})}^2+{(3{\mathrm{\η}}_{21}-{\mathrm{\η}}_{03})}^2;\\ I_4={({\mathrm{\η}}_{30}+{\mathrm{\η}}_{12})}^2+{({\mathrm{\η}}_{21}+{\mathrm{\η}}_{03})}^2;\\ I_5=({\mathrm{\η}}_{30}-3{\mathrm{\η}}_{12})({\mathrm{\η}}_{30}+{\mathrm{\η}}_{12})({\left({\mathrm{\η}}_{30}+{\mathrm{\η}}_{12}\right)}^2-3{\left({\mathrm{\η}}_{21}+{\mathrm{\η}}_{03}\right)}^2)\\ +(3{\mathrm{\η}}_{21}-{\mathrm{\η}}_{03})({\mathrm{\η}}_{21}+{\mathrm{\η}}_{03})(3{\left({\mathrm{\η}}_{30}+{\mathrm{\η}}_{12}\right)}^2-{\left({\mathrm{\η}}_{03}+{\mathrm{\η}}_{12}\right)}^2);\\ I_6=({\mathrm{\η}}_{20}-{\mathrm{\η}}_{02}){({\mathrm{\η}}_{30}+{\mathrm{\η}}_{12})}^2-{({\mathrm{\η}}_{03}+{\mathrm{\η}}_{21})}^2+4{\mathrm{\η}}_{11}({\mathrm{\η}}_{30}+{\mathrm{\η}}_{12})({\mathrm{\η}}_{03}+{\mathrm{\η}}_{12});\\ I_7=(3{\mathrm{\η}}_{21}+{\mathrm{\η}}_{03})({\mathrm{\η}}_{30}+{\mathrm{\η}}_{12}){({\mathrm{\η}}_{12}+{\mathrm{\η}}_{30})}^2-3{({\mathrm{\η}}_{21}+{\mathrm{\η}}_{03})}^2\\ +(3{\mathrm{\η}}_{12}-{\mathrm{\η}}_{30})({\mathrm{\η}}_{21}+{\mathrm{\η}}_{03})(3{\left({\mathrm{\η}}_{30}+{\mathrm{\η}}_{12}\right)}^2-{\left({\mathrm{\η}}_{21}+{\mathrm{\η}}_{03}\right)}^2);\end{array}---\left(7\right)$

2-3 area Difference Calculation: the area S of the standard abnormity stamping parts that step 2-1 is obtained₀With abnormity to be detected punching press The area S of part_nDiffer from, obtain the area difference result Δ S of abnormity stamping parts to be detected and standard abnormity stamping parts_n, it may be assumed that

ΔS_n=| S₀-S_n|； (8)

The calculating of 2-4 outline rate: the Hu not bending moment I obtained according to step 2-2_i(1≤i≤7), mark is passed judgment in definition Accurate:

$I(A,B)=\underset{i=1}{\overset{7}{\Σ}}|\frac{1}{{m_i}^A}-\frac{1}{{m_i}^B}|;---\left(9\right)$

Wherein, m_i ^A、m_i ^BIt is defined as:

m_i ^A=sign (I_i ^A)·log|I_i ^A|； (10)

m_i ^B=sign (I_i ^B)·log|I_i ^B|； (11)

Ask for the matching rate (in units of percentage ratio) of standard workpiece and workpiece to be detected:

I=100-I (A, B) × 100； (12)

In formula, A, B refer to standard workpiece and workpiece to be checked respectively；

3rd step, detects in real time

3-1 image real-time acquisition: detecting system start after, control part control carrying workpiece transparent rotating circular disk 8 with Constant speed rotates counterclockwise, when workpiece motion s is to when having between the correlation position of sensor transmitter and transducer receivers, Collection signal is transferred to control part by sensor, and then the transparent rotating circular disk controlling carrying workpiece stops 1s, it is thus possible to right Workpiece carries out stationary picture；Control part simultaneously is signaled to smart camera 3, utilizes smart camera external trigger pattern to trigger intelligence Energy camera Real-time Collection workpiece image, calls template image to carry out next step image procossing；

3-2 image procossing: after image acquisition terminates, carries out image procossing by computer 12, by the profile of step 2-4 Matching rate judgment criteria, the area difference result in conjunction with step 2-3 sets threshold value, carries out matching detection, it is judged that workpiece whether Qualified, and feed back to control part；

3-3 workpiece is classified: after image procossing has judged that workpiece is the most qualified, next by outside grasping mechanism to work Part captures, and qualified storehouse put into by qualified workpiece, otherwise puts into recovery storehouse, thus completes the detection of whole workpiece, defect.

Embodiment illustrated in fig. 1 shows, the present invention special-shaped stamping parts profile defects vision detection system based on coupling is made Assembling test platform hardware composition include smart camera the 3, first support the 1, second support 2, computer 12, control portion Point, transparent rotating circular disk 8, motor 7, sensor and backlight 4, control part and control the rotation of motor 7 and smart camera 3 Trigger；Being collectively formed visual system by smart camera the 3, first support 1 and computer 12, described smart camera 3 is by first Frame 1 is fixed on the top of transparent rotating circular disk 8, and smart camera 3 is connected with computer 12 and control part by order wire simultaneously, Computer 12 with control part is connected, be used for show smart camera 3 gather and process after image；Described transparent rotating circular disk 8 Being connected with motor 7, described backlight 4 is positioned at the underface of smart camera 3, and is in the lower section of transparent rotating circular disk 8；Described Sensor includes sensor transmitter 5 and transducer receivers 6, and transducer receivers 6 is oppositely arranged with sensor transmitter 5, And in the same horizontal line, sensor transmitter 5 is connected with the second support 2, fixes position and complete the accurate transmission of signal, Transducer receivers 6 is fixed on the first support 1 bottom, and transducer receivers 6 connects with control part with sensor transmitter 5 simultaneously Connecing, the two has coordinated the collection of stamping parts position signalling.

The control part of the present invention drives transparent rotating circular disk 8 to rotate by controlling motor 7, and transparent rotating circular disk 8 holds Carry abnormity stamping parts 10, and then realize the direction of transfer of stamping parts 10 and the control of speed, simultaneously sensor transmitter 5 with The position signalling of the stamping parts 10 that transducer receivers 6 assisted acquisition arrives, is transferred to control part, controls part and controls intelligence phase The triggering collection of machine, the image transmitting that smart camera 3 collects, to computer 12, is carried out image procossing, image by computer 12 After having processed, control part then controls outside grasping mechanism and classifies the special-shaped stamping parts after having detected.Backlight 4 can provide illumination so that the contrast between target and background is obvious, and the image of collection is apparent accurately.

Embodiment illustrated in fig. 2 shows, the present invention special-shaped stamping parts profile defects vision detection system based on coupling is in intelligence The schematic diagram of special-shaped stamping parts 10 when being in diverse location under camera 3 visual field, can be gathered, by diagram it can be seen that due to different There is certain thickness in shape stamping parts, the various location under stamping parts is in smart camera 3 visual field, and the angle of visual field can occur one Fixed change, therefore can produce some impacts, and then affect final Detection results the image gathered.

Fig. 3 is the structural representation of the special-shaped stamping parts of an embodiment of the present invention, is to test targeted destination object, Embodiment illustrated in fig. 4 shows, due to the fact that sensor transmitter 5 and the existence of transducer receivers 6, when abnormity stamping parts 10 Arrive under the visual field of smart camera 3, and when being in sensing station, gather signal and be transferred to control part, control to trigger Smart camera 3 gathers image, and such mode makes stamping parts consistent in the visual field general scope of smart camera 3, will not occur The biggest skew, and then decrease the impact on detection profile defects of the stamping parts location, improve the reliability of detection.

Embodiment illustrated in fig. 5 shows, the present invention special-shaped stamping parts profile defects vision detection system based on coupling, is carrying Adding very noisy before contouring, the outline effect schematic diagram obtained, in figure, black splotch is illustrated as noise, as can be seen from the figure Still can extract the contour images of abnormity stamping parts after adding very noisy exactly, the algorithm of contours extract of the present invention is independent of In the gradient of image, having the strongest antinoise, capacity of resisting disturbance, strong robustness, contours extract is more accurate.

The present invention according to the size of workpiece, and can account for the size in the smart camera visual field, adjust camera lens size and The object distance of camera.

Embodiment

The present embodiment special-shaped stamping parts profile defects vision detection system based on coupling includes smart camera 3, first Frame the 1, second support 2, computer 12, control part, transparent rotating circular disk 8, motor 7, sensor and backlight 4, control portion The rotation of sub-control motor 7 and the triggering of smart camera 3；Collectively formed regarded by smart camera the 3, first support 1 and computer 12 Vision system, described smart camera 3 is fixed on the top of transparent rotating circular disk 8 by the first support 1, and smart camera 3 is by communication Line simultaneously with computer 12 and control part and is connected, computer 12 with control partly to be connected, be used for showing smart camera 3 collection with Image after process；Described transparent rotating circular disk 8 is connected with motor 7, and described backlight 4 is positioned at the underface of smart camera 3, And it is in the lower section of transparent rotating circular disk 8；Described sensor includes sensor transmitter 5 and transducer receivers 6, and sensor connects Receive device 6 to be oppositely arranged with sensor transmitter 5, and in the same horizontal line, sensor transmitter 5 be connected with the second support 2, Fixing position and complete the accurate transmission of signal, transducer receivers 6 is fixed on the first support 1 bottom, transducer receivers 6 with Sensor transmitter 5 is connected with control part simultaneously, and the two has coordinated the collection of stamping parts position signalling.

The present embodiment special-shaped stamping parts profile defects visible detection method based on coupling, uses above-mentioned detecting system, Comprising the concrete steps that of the method:

The first step, image procossing

1-1 Image Acquisition: obtain contrast significantly abnormity stamping parts image by smart camera 3；

1-2 image threshold adaptivenon-uniform sampling: on the basis of step 1-1, calculates each pixel for the image collected The weighted mean in point 5 × 5 regions around, deducting a constant obtains adaptive threshold, and each pixel pixel value is more than threshold Value, it is defined as the available point of interesting target object abnormity stamping parts in image；

Seeking of 1-3 profile takes: on the basis of step 1-2, has interesting target object abnormity stamping parts in image Effect point carries out the continuous evolution of curve, arranges iterations, can obtain effective contour curve C (x, y) figure of abnormity stamping parts Picture；

Second step, image information analysis

The calculating of 2-1 area: the contour curve C that obtains according to step 1-3 (x, y) is divided into two parts by image-region, A part is that (x, y), another part is that (x, y), to obtain for contour curve perimeter outC to contour curve interior zone inC (x, y) is integrated contour curve interior zone inC, i.e. can obtain characterizing the area information of contoured interior area size, this step Suddenly the area S of standard abnormity stamping parts can be calculated₀Area S with special-shaped stamping parts to be detected_n, wherein, S_nRefer to The area of n workpiece to be detected, n >=1；

The computing formula of described contoured interior region area is formula (1):

S=∫_inC(x,y)u(x,y)dxdy； (1)

Wherein, (x y) is the contour curve density function that extracts after threshold adaptive dividing processing of image to u；

The calculating of 2-2 profile Hu not bending moment: (x, y), by owning on profile for the contour curve C obtained for step 1-3 Point carries out integral operation, obtains profile Hu not bending moment, it is assumed that the some number on profile is N, then can obtain the p+q rank of profile Square, such as formula:

$m_{pq}=\underset{i=1}{\overset{N}{\Σ}}{x}^p{y}^{q}u(x,y);---\left(2\right)$

Profile p+q rank central moment is obtained by formula (3),

${\mathrm{\μ}}_{pq}=\underset{i=1}{\overset{N}{\Σ}}{(x-\stackrel{\‾}{x})}^p{(y-\stackrel{\‾}{y})}^{q}u(x,y);---\left(3\right)$

In formula, the square on p correspondence x dimension, the square on q correspondence y-dimension,WithRepresent the center of gravity of profile:

$\stackrel{\‾}{x}=m_{10}/m_{00};---\left(4\right)$

$\stackrel{\‾}{y}=m_{01}/m_{00};---\left(5\right)$

Normalized p+q central moment η_pqIt is defined as:

${\mathrm{\η}}_{pq}={\mathrm{\μ}}_{pq}/\left({\mathrm{\μ}}_{00}^{\mathrm{\ρ}}\right);---\left(6\right)$

Wherein, ρ=(p+q)/2+1；

Hu not bending moment is the linear combination of the normalization central moment obtained by formula (6), can obtain profile by formula (3) Second order and third central moment, be brought into formula (6) and obtain second order three rank normalization central moment, by second order and three normalization centers, rank Square seven Hu not bending moment I of structure₁～I₇, the formula of concrete Hu not bending moment is formula (7), and Hu not bending moment can under the conditions of consecutive image Keep translation, scaling and invariable rotary；:

$\begin{array}{c}{I}_1={\mathrm{\η}}_{20}+{\mathrm{\η}}_{02};\\ I_2={({\mathrm{\η}}_{20}-{\mathrm{\η}}_{02})}^2+4{\mathrm{\η}}_{11}^2;\\ I_3={({\mathrm{\η}}_{30}-{\mathrm{\η}}_{12})}^2+{(3{\mathrm{\η}}_{21}-{\mathrm{\η}}_{03})}^2;\\ I_4={({\mathrm{\η}}_{30}+{\mathrm{\η}}_{12})}^2+{({\mathrm{\η}}_{21}+{\mathrm{\η}}_{03})}^2;\\ I_5=({\mathrm{\η}}_{30}-3{\mathrm{\η}}_{12})({\mathrm{\η}}_{30}+{\mathrm{\η}}_{12})({\left({\mathrm{\η}}_{30}+{\mathrm{\η}}_{12}\right)}^2-3{\left({\mathrm{\η}}_{21}+{\mathrm{\η}}_{03}\right)}^2)\\ +(3{\mathrm{\η}}_{21}-{\mathrm{\η}}_{03})({\mathrm{\η}}_{21}+{\mathrm{\η}}_{03})(3{\left({\mathrm{\η}}_{30}+{\mathrm{\η}}_{12}\right)}^2-{\left({\mathrm{\η}}_{03}+{\mathrm{\η}}_{12}\right)}^2);\\ I_6=({\mathrm{\η}}_{20}-{\mathrm{\η}}_{02}){({\mathrm{\η}}_{30}+{\mathrm{\η}}_{12})}^2-{({\mathrm{\η}}_{03}+{\mathrm{\η}}_{21})}^2+4{\mathrm{\η}}_{11}({\mathrm{\η}}_{30}+{\mathrm{\η}}_{12})({\mathrm{\η}}_{03}+{\mathrm{\η}}_{12});\\ I_7=(3{\mathrm{\η}}_{21}+{\mathrm{\η}}_{03})({\mathrm{\η}}_{30}+{\mathrm{\η}}_{12}){({\mathrm{\η}}_{12}+{\mathrm{\η}}_{30})}^2-3{({\mathrm{\η}}_{21}+{\mathrm{\η}}_{03})}^2\\ +(3{\mathrm{\η}}_{12}-{\mathrm{\η}}_{30})({\mathrm{\η}}_{21}+{\mathrm{\η}}_{03})(3{\left({\mathrm{\η}}_{30}+{\mathrm{\η}}_{12}\right)}^2-{\left({\mathrm{\η}}_{21}+{\mathrm{\η}}_{03}\right)}^2);\end{array}---\left(7\right)$

2-3 area Difference Calculation: the area S of the standard abnormity stamping parts that step 2-1 is obtained_b0With abnormity to be detected punching press The area S of part_nDiffer from, obtain the area difference result Δ S of abnormity stamping parts to be detected and standard abnormity stamping parts_bn, it may be assumed that

ΔS_bn=| S_b0-S_n|； (8)

Area difference result Δ S when the n-th abnormity stamping parts to be detected with b standard abnormity stamping parts_bnIn, as long as depositing At a parameter value in threshold range, i.e. Δ S_bn＜ yuzhi, decides that this abnormity stamping parts to be detected is qualified products；

Wherein, S_b0Refer to the area of b standard abnormity stamping parts, S_nRefer to the area of the n-th workpiece to be detected, this In embodiment, b is 4, and n is 5；

The calculating of 2-4 outline rate: the Hu not bending moment I obtained according to step 2-2_i(1≤i≤7), mark is passed judgment in definition Accurate:

$I(A,B)=\underset{i=1}{\overset{7}{\Σ}}|\frac{1}{{m_i}^A}-\frac{1}{{m_i}^B}|;---\left(9\right)$

Wherein, m_i ^A、m_i ^BIt is defined as:

m_i ^A=sign (I_i ^A)·log|I_i ^A|； (10)

m_i ^B=sign (I_i ^B)·log|I_i ^B|； (11)

Ask for the matching rate (in units of percentage ratio) of standard workpiece and workpiece to be detected:

I_kl=100-I (A_k,B_l)×100； (12)

In formula, A_k, B_lRefer to kth standard workpiece and l workpiece to be checked respectively；

3rd step, detects in real time

3-1 image real-time acquisition: detecting system start after, control part control carrying workpiece transparent rotating circular disk 8 with Constant speed rotates counterclockwise, when workpiece motion s is to when having between the correlation position of sensor transmitter and transducer receivers, Collection signal is transferred to control part by sensor, and then the transparent rotating circular disk controlling carrying workpiece stops 1s, it is thus possible to right Workpiece carries out stationary picture；Control part simultaneously is signaled to smart camera 3, utilizes smart camera external trigger pattern to trigger intelligence Energy camera Real-time Collection workpiece image, calls template image to carry out next step image procossing；

3-2 image procossing: after image acquisition terminates, carries out image procossing by computer 12, by the profile of step 2-4 Matching rate judgment criteria, the area difference result in conjunction with step 2-3 sets threshold value, carries out matching detection, it is judged that workpiece whether Qualified, and feed back to control part；

3-3 workpiece is classified: after image procossing has judged that workpiece is the most qualified, next by outside grasping mechanism to work Part captures, and qualified storehouse put into by qualified workpiece, otherwise puts into recovery storehouse, thus completes the detection of whole workpiece, defect.

The present embodiment connects device on request, and selecting abnormity stamping parts thickness is 4mm, and size is 43mm × 45mm (seeing Fig. 3).Smart camera distance transparent rotating circular disk 300mm, backlight 4 is positioned at the underface of smart camera 3, and is in The lower section of transparent rotating circular disk 8, apart from bright rotating circular disk about 20mm,

In use, be first turned on smart camera power supply, with detection during under equivalent environment, gather smart camera and regard The standard abnormity stamping parts image of wild lower different positions and pose, as template, does premise for ensuing matching detection and prepares.According to upper The method step stated detects.

Table 1 be abnormity stamping parts 10 occur in the scope of smart camera 3 unanimous on the whole in the case of, to same standard Abnormity stamping parts rotates, and obtains standard abnormity stamping parts and is in the situation of different positions and pose, in the case of different positions and pose Standard abnormity stamping parts carries out image acquisition, the area of the different positions and pose collected.This example is tested by four poses, Obtaining area result, there is certain difference in the area that different pose presentation are corresponding as can be seen from Table 1, if standard abnormity punching press Part is different from not having defective abnormity stamping parts pose, and area difference is likely to can be more than standard abnormity stamping parts in some cases With the area difference of defective abnormity stamping parts, erroneous judgement can be caused, due to the requirement of industrial detection precision, it is impossible to separately through Set the threshold value of area difference coupling as the standard judging defect problem the most accurately.

Table 2 be abnormity stamping parts 10 occur in the scope of smart camera 3 unanimous on the whole in the case of, for standard abnormity Stamping parts different positions and pose carries out image acquisition, the Hu obtained not bending moment parametric results.Result shows that different pose presentation is corresponding Hu not bending moment parameter I₁～I₇Change is little, demonstrates again that the robustness of Hu not bending moment, decreases pose different for testing result Impact, in order to identify exactly abnormity stamping parts 10 defect, matching result can by obtain each to be detected abnormity punching The Hu square parameter of casting die compares with each standard abnormity stamping parts 10 and judges, and combines each abnormity punching to be detected The area features as shown in table 1 of casting die carries out area difference with each mark abnormity stamping parts and mates.

Table 3 is the application present invention special-shaped stamping parts profile defects vision detection system based on coupling and method, obtains Two qualified workpiece and three defective workpiece respectively with the area difference result of standard abnormity stamping parts.This table selects four Individual standard abnormity stamping parts, as area difference template, is designated as respectively: template 1, template 2, template 3 and template 4, can by result To find out that qualified workpiece and defective workpiece are for multiple standard workpiece area difference result Δ S_bn, wherein b is standard workpiece Number, n is the number of workpiece to be detected, it can be seen that the area difference result of certified products and standard abnormity stamping parts template, With the area difference result gap of defective abnormity stamping parts and standard abnormity stamping parts template or big, tie according to difference Fruit sets area difference matching threshold.Each abnormity stamping parts to be detected, relative to different templates, area difference result is not With, select area difference result minima Δ S_minIf, less than threshold value, prove abnormity stamping parts to be detected at least with one of them Template matching, it is possible to tentatively judge that abnormity stamping parts to be detected is defective as not having, so can tentatively judge work to be detected The defect situation of part.

Table 4 is the application present invention special-shaped stamping parts profile defects vision detection system based on coupling and method, obtains Two qualified workpiece and three defective workpiece Hu matching results.Result shows qualified work piece match rate I_bjThe highest, 99% Above, this joins rate (only up to about 90%) with defective workpiece significantly differentiation, sets work piece match rate threshold value, only That wants abnormity stamping parts and each standard abnormity stamping parts template to be detected joins in rate, has just can sentence more than the threshold value set The matching rate I that disconnected abnormity stamping parts to be detected is defective for not having, maximum in human-computer interaction interface display matching rate_max(the most final Matching rate), in conjunction with result and the judgement of table 3, the matching rate I that display is maximum_max, it is possible to accurately detect and account for whole workpiece face The defect of long-pending more than 0.5%, improves accuracy of detection, it is possible to meet the accuracy requirement of defects detection in actual field.

Table 1

Table 2

Table 3

Table 4

Claims (3)

1. a special-shaped stamping parts profile defects vision detection system based on coupling, it is characterised in that this detecting system includes intelligence Energy camera, the first support, the second support, computer, control part, transparent rotating circular disk, motor, sensor and backlight, Control part and control rotation and the triggering of smart camera of motor；Collectively formed regarded by smart camera, the first support and computer Vision system, described smart camera is fixed on the top of transparent rotating circular disk by the first support, and smart camera is same by order wire Time with computer and control part and is connected, computer is partly connected with control, and described transparent rotating circular disk is connected with motor, described Backlight is positioned at the underface of smart camera, and is in the lower section of transparent rotating circular disk；Described sensor includes that sensor is sent out Emitter and transducer receivers, transducer receivers is oppositely arranged with sensor transmitter, and in the same horizontal line, sensing Device emitter and the second support are connected, and transducer receivers is fixed on the first support bottom, and transducer receivers is sent out with sensor Emitter is connected with control part simultaneously.

Special-shaped stamping parts profile defects vision detection system based on coupling the most according to claim 1, it is characterised in that The thickness of described abnormity stamping parts is 3～5mm, and stamping parts a length of 40～60mm, width is 15～55mm.

3. a special-shaped stamping parts profile defects visible detection method based on coupling, uses the detection described in claim 1 or 2 System, the method utilizes the standard extracted and the profile information of abnormity stamping parts to be detected, according to utilizing standard workpiece and having scarce Fall into area discrepancy that workpiece exists and there is rotation, translate, scale the difference of the profile not bending moment of invariance, contrasting to be detected different Shape stamping parts, the area of standard abnormity stamping parts and profile not bending moment, be calculated the coupling of difference in areas and profile respectively Rate, then set threshold value respectively, it is judged that abnormity stamping parts whether existing defects, comprising the concrete steps that of the method:

The first step, image procossing

1-1 Image Acquisition: obtain contrast significantly abnormity stamping parts image by smart camera；

1-2 image threshold adaptivenon-uniform sampling: on the basis of step 1-1, the image for collecting calculates each pixel week Enclosing the weighted mean in 5 × 5 regions, deduct a constant and obtain adaptive threshold, each pixel pixel value is more than threshold value, It is defined as the available point of interesting target object abnormity stamping parts in image；

Seeking of 1-3 profile takes: on the basis of step 1-2, to the available point of interesting target object abnormity stamping parts in image Carry out the continuous evolution of curve, iterations is set, effective contour curve C (x, y) image of abnormity stamping parts can be obtained；

Second step, image information analysis

The calculating of 2-1 area: (x, y) is divided into two parts by image-region, one according to the contour curve C that step 1-3 obtains Dividing is that (x, y), another part is that (x, y), to the profile obtained for contour curve perimeter outC to contour curve interior zone inC (x, y) is integrated curvilinear inner region inC, i.e. can obtain characterizing the area information of contoured interior area size, and this step can It is calculated the area S of standard abnormity stamping parts₀Area S with special-shaped stamping parts to be detected_n, wherein, S_nRefer to n-th The area of workpiece to be detected, n >=1；

The calculating of 2-2 profile Hu not bending moment: (all on profile y), are clicked on by x for the contour curve C that obtains for step 1-3 Row integral operation, obtains profile Hu not bending moment, it is assumed that the some number on profile is N, obtains center, profile p+q rank by formula (3) Square,

${\mathrm{\μ}}_{pq}=\underset{i=1}{\overset{N}{\Σ}}{(x-\stackrel{\‾}{x})}^p{(y-\stackrel{\‾}{y})}^{q}f(x,y);---\left(3\right)$

In formula, the square on p correspondence x dimension, the square on q correspondence y-dimension,WithRepresent the center of gravity of profile:

Normalized p+q central moment η_pqIt is defined as:

${\mathrm{\η}}_{pq}={\mathrm{\μ}}_{pq}/\left({\mathrm{\μ}}_{00}^{\mathrm{\ρ}}\right);---\left(6\right)$

Wherein, ρ=(p+q)/2+1；

Hu not bending moment is the linear combination of the normalization central moment obtained by formula (6), can be obtained the second order of profile by formula (3) And third central moment, it is brought into formula (6) and obtains second order three rank normalization central moment, by second order and three rank normalization central moment structures Make seven Hu not bending moment I₁～I₇, the formula of concrete Hu not bending moment is formula (7), and Hu not bending moment can keep under the conditions of consecutive image Translation, scaling and invariable rotary；:

$\begin{array}{c}{I}_1={\mathrm{\η}}_{20}+{\mathrm{\η}}_{02};\\ I_2={({\mathrm{\η}}_{20}-{\mathrm{\η}}_{02})}^2+4{\mathrm{\η}}_{11}^2;\\ I_3={({\mathrm{\η}}_{30}-{\mathrm{\η}}_{12})}^2+{(3{\mathrm{\η}}_{21}-{\mathrm{\η}}_{03})}^2;\\ I_4={({\mathrm{\η}}_{30}+{\mathrm{\η}}_{12})}^2+{({\mathrm{\η}}_{21}+{\mathrm{\η}}_{03})}^2;\\ I_5=({\mathrm{\η}}_{30}-3{\mathrm{\η}}_{12})({\mathrm{\η}}_{30}+{\mathrm{\η}}_{12})({\left({\mathrm{\η}}_{30}+{\mathrm{\η}}_{12}\right)}^2-3{\left({\mathrm{\η}}_{21}+{\mathrm{\η}}_{03}\right)}^2)\\ +(3{\mathrm{\η}}_{21}-{\mathrm{\η}}_{03})({\mathrm{\η}}_{21}+{\mathrm{\η}}_{03})(3{\left({\mathrm{\η}}_{30}+{\mathrm{\η}}_{12}\right)}^2-{\left({\mathrm{\η}}_{03}+{\mathrm{\η}}_{12}\right)}^2);\\ I_6=({\mathrm{\η}}_{20}-{\mathrm{\η}}_{02}){({\mathrm{\η}}_{30}+{\mathrm{\η}}_{12})}^2-{({\mathrm{\η}}_{03}+{\mathrm{\η}}_{21})}^2+4{\mathrm{\η}}_{11}({\mathrm{\η}}_{30}+{\mathrm{\η}}_{12})({\mathrm{\η}}_{03}+{\mathrm{\η}}_{21});\\ I_7=(3{\mathrm{\η}}_{21}+{\mathrm{\η}}_{03})({\mathrm{\η}}_{30}+{\mathrm{\η}}_{12}){({\mathrm{\η}}_{12}+{\mathrm{\η}}_{30})}^2-3{({\mathrm{\η}}_{21}+{\mathrm{\η}}_{03})}^2\\ +(3{\mathrm{\η}}_{12}-{\mathrm{\η}}_{30})({\mathrm{\η}}_{21}+{\mathrm{\η}}_{03})(3{\left({\mathrm{\η}}_{30}+{\mathrm{\η}}_{12}\right)}^2-{\left({\mathrm{\η}}_{21}+{\mathrm{\η}}_{03}\right)}^2);\end{array}---\left(7\right)$

2-3 area Difference Calculation: the area S of the standard abnormity stamping parts that step 2-1 is obtained₀With abnormity stamping parts to be detected Area S_nDiffer from, obtain the area difference result Δ S of abnormity stamping parts to be detected and standard abnormity stamping parts_n, i.e. Δ S_n=| S₀- S_n|；

The calculating of 2-4 outline rate: the Hu not bending moment I obtained according to step 2-2_i(1≤i≤7), definition judgment criteria:

$I(A,B)=\underset{i=1}{\overset{7}{\Σ}}|\frac{1}{{m_i}^A}-\frac{1}{{m_i}^B}|;---\left(9\right)$

Wherein, m_i ^A、m_i ^BIt is defined as:

m_i ^A=sign (I_i ^A)·log|I_i ^A|； (10)

m_i ^B=sign (I_i ^B)·log|I_i ^B|； (11)

Ask for the matching rate (in units of percentage ratio) of standard workpiece and workpiece to be detected:

I=100-I (A, B) × 100； (12)

In formula, A, B refer to standard workpiece and workpiece to be checked respectively；

3rd step, detects in real time

3-1 image real-time acquisition: after detecting system starts, controls part and controls the transparent rotating circular disk of carrying workpiece with constant speed Degree rotates counterclockwise, when workpiece motion s is to when having between the correlation position of sensor transmitter and transducer receivers, and sensor Collection signal is transferred to control part, and then the transparent rotating circular disk controlling carrying workpiece stops 1s, it is thus possible to enter workpiece Row stationary picture；Control part simultaneously is signaled to smart camera, utilizes smart camera external trigger pattern to trigger smart camera Real-time Collection workpiece image, calls template image to carry out next step image procossing；

3-2 image procossing: after image acquisition terminates, carries out image procossing by computer, by the outline rate of step 2-4 Judgment criteria, the area difference result in conjunction with step 2-3 sets threshold value, carries out matching detection, it is judged that workpiece is the most qualified, and Feed back to control part；

3-3 workpiece is classified: after image procossing has judged that workpiece is the most qualified, is next entered workpiece by outside grasping mechanism Row captures, and qualified storehouse put into by qualified workpiece, otherwise puts into recovery storehouse, thus completes the detection of whole workpiece, defect.