The welding line groove that is based on 2 thought are little commutation and pattern recognition identifies

5MPa) , large requirements (way of 1 420 Mm) develops, of welding line of link of this pair of pipeline solder raised taller requirement, of pipeline link welding line solder make the crucial process that restricts whole project quality and construction cycle. Normal > in view of worldwide inside the huge market with potential construction of oil gas cop, to improve welding line quality, reduce worker labor intensity, implementation conduit builds fieldwork automation, shorten construction is periodic, a lot of more domestic and international the development development that well-known pipeline company has crossing conduit to solder automatically the technology reachs device. Normal > is well-known, what pipeline link seams is automatic dogging is the premise that implementation solders automatically, and the extraction that annulus seams information is the requirement that implementation dogs automatically, undertook in the light of this one problem pipeline link seams groove identifying research. Normal > is opposite at present respect of processing of welding line image uses the method that is based on mutation of image gray scale generally. This kind of method does not apply to mutation of groove brim gray scale surface of not apparent, workpiece is put in the circumstance such as interference of a lot of noise. Pattern recognition is a new course that rapid development rises at the beginning of 60 time, belong to information, control and the category with scientific system. Rise quickly as what the development of technology of large scale integration and computer sex price compare, no matter be in theoretic, still go up in application, pattern recognition technology has remarkable development, promoted the development of the academic direction such as pattern recognition and new technology. Mode is become to take the sheet from world finite share by understanding same this by measured value integrated; Pattern recognition tries to decide the category attribute of an example namely, namely some example vest in the certain model in many models [1] . Pattern recognition system should finish mode to collect, the feature is extracted / the function such as choice and classification. Seam groove to annulus for, although different pipeline link seams the grain feature of groove to differ, but the grain feature that seams groove with a annulus keeps basically consistent, because this is aimed at each annulus to seam groove to build one corresponding pattern plate, use pattern plate to match methodological identifying annulus to seam groove. Use method of 2 thought are little commutation to undertake handling to initiative groove image above all, undertake to small commutation result 2 values change processing, get the grain information of image, determine more appropriate small commutation measure through optimizing algorithm next, the 2 values that determine outcome of small commutation below this measure change image to be pattern plate; Small shift that makes same measure fall to follow-up image reachs 2 values to change processing, the follow-up after using pattern plate to change processing in 2 values undertakes pattern plate matchs computation on image, determine position of follow-up image groove, output data of center of follow-up image groove finally. Say wraparoundly, crack ring groove identifying cent to undertake for 4 paces. Normal > (1) the collection of groove image. Normal > (2) build annulus to seam groove pattern plate. Normal > (3) seam groove to undertake discriminating to annulus. Normal > (4) data of position of output groove center. Normal >1 groove image collects Normal >normal > in view of soldering robot in all sorts of vision sensor, CCD sensor has property reliable, get image clarity intuitionistic and use convenient wait for an advantage. The article uses a face blast CCD will get picture of welding line groove. Collect card to will get signal of number of imitate image translate into to send the computer through Matrox Meteor-II/Standard image. Graph 1 in be opposite for the welding torch below initiative condition good when gotten ring seams groove image. Normal Align=center> pursues image of groove of welding line of 1 one frame 2 build pattern plate of welding line groove 2.

1 annulus seams the initiative condition Normal Align=left> seams groove according to pipeline link to assemble a characteristic like the head, can determine the following and initiative condition: Normal > (1) direction of welding line groove is perpendicular direction basically. Normal > (2) welding torch is to solder below initiative condition center, CCD photographs center resembling a head and welding torch center to keep consistent. Normal > (3) width of welding line groove is foregone. Normal >2.

The 2 grain information Normal > image are to be in dimensional position camber is local those who change, these composition are not similar to any Fourier base function, and their commutation coefficient also is not compact distributing. This makes the commutation method such as Fourier commutation includes transient state in the analysis or local when the signal that spends part and image, cannot get optimal express. For this, mathematician and engineers are developed use finite width base the method that function has alternating. These base function is not only on frequency and be change locally, the wave that they are finite width is called small. The commutation that is based on them is called small to alternate [2] . Can saying as a result of function of base B batten is coming true to software or hardware is more effective the simplest function that has Xiaozhi to maintain, use B batten small wave to undertake information of groove image grain is extracted so. Take measure to be 2m following 3 direction are little type place are shown (1) in type, , , be N for the frequency respectively the first, 2, 3 direction are little M -- 2 footage spend factor -- the N that measure is 2m 2. Correspondence is at 3 direction in frequency domain are little definition (2) Normal > type -- N second the Fourier commutation Normal >wx, Wy -- it is correspondence respectively the solid frequency Normal > , the B batten function that is N for the frequency respectively is corresponding the first, 2, 3 direction are little frequency domain expresses Normal >G(1) to take batten of kind of Canny functor B one Wei Xiaobo FIR filters the transfer function of coefficient G(1) , g(2) to take batten of kind of LoG functor B one Wei Xiaobo FIR filters the transfer function of coefficient G(2) . G(1) takes: G1= - 1, g2=1, gk=0 K ≠ 1, 2g(2) takes: G1=1, g2= - 2, g3=1, gk=0k ≠ 1, 2, 3normal > calculates what spend Z of alignment {2m}m ∈ along 2 footage is local slant guide of heft recursive and algorithmic for: (3) Normal > type -- what undertake to image matrix F measure is 2m is low connect flowing, it is respectively use the first, 2, small shift that to image matrix F has 3 direction are little measure is 2m expresses to use image matrix F flowing below 2m-1 measure after. H of coefficient of flowing filter wave takes: H2=0.

0625, h3=0.

25, h4=0.

375, h5=0.

25, h6=0.

0625, hk=0 (k ≠ 2, 3, 4, 5, 6) . Symbolic D filters on behalf of Dirac coefficient, this coefficient is in origin 1, in it is elsewhere 0. Normal > is taken photograph the primitive picture that gets like head and image card through CCD. Normal > is perpendicular direction as a result of groove, be opposite with the first direction is little graph 1 undertake small wave alternates, get is as follows as a result pursueing what showing. Normal Align=center>(a) is small measure is 22 Normal Align=center> pursues 2 use different measure small wave to the graph 1 undertake Normal > undertakes to small change result 2 values change processing (4) (5) type is medium, -- undertake to F with the first direction is little measure is 21 reach small commutation of 22 T1, T2 -- the 2 values image of correspondence of graph 2 A, B -- the first, j) the value T1 that the dot is in, T2 -- corresponding the threshold value F at graph 2 A, B -- represent an attempt 1 among them T1, T2 is passed respectively after in be being reached to commutation matrix, each data seeks absolute value, be added and beg get on average. Earning following plan institute show 2 values picture Normal Align=center> pursues 3 correspondence at the graph the 2 values of 2 change image 2.

Normal > by the graph 3 visible, the 2 values that are not result of small commutation below all measure change image to suit to serve as pattern plate. According to initiative condition the article builds the target function that one pattern plate chooses: Tt(x1 of Min Vm- ∑ , x2)+x3 × 0.

2 × Vm (6)s.

Tt(x1 of T Vm- ∑ , x2) ≤ 0.

2 × Vmnormal >x1=Hcnormal >x2 ≤ Vmnormal >x2 ≥ -- Normal >Tt -- Normal >x1 -- Normal >x2 -- Normal >x3 -- frequency Normal >Hc -- Normal > undertakes to small commutation result 2 values change processing to calculate 2 values are 1 elephant on picture center line element the sum is 1 elephant element and ＞ 0.

8 × Vm? Undertake the small commutation M=1 of 2m measure to initiative image, x3=0x3=x3+1 is affirmatory and current 2 values image is pattern plate M=m+1NYnormal Align=center>normal > target function is the 80% above that achieve image length in length of line of center of 2 values image least second small commutation, in order to reduce computational time. If the graph is shown 4 times,optimize algorithmic flow chart. If the graph is shown 5 times,obtain pattern plate finally. To raise pattern recognition precision, did not take whole 2 values image serves as pattern plate, take however among them a paragraph, take pattern plate dimension to be 100 × 280 (like element) . Normal Align=center>normal >3 use pattern plate matchs methodological identifying annulus to seam use below groove above affirmatory pattern plate matchs groove of algorithm identifying welding line through pattern plate. Algorithmic flow chart is shown 6 times like the graph. Computation is shown 7 times like the graph as a result, by the graph 7 visible, one peak value exists in the graph, criterion the image area of correspondence of this peak value is this paragraphs groove area in image. 4 positions data outputs Normal >4.

The deviation of welding torch of 1 tie condition and way of level of groove center position measures nonexistent mutation, the absolute value of quantity of groove center deviation that gets picture twice continuously is less than L (decide according to actual condition) . 4.

If the graph is shown 8 times,2 calm mark and data of welding torch position are outputted, be in with picture center (position of way of standard of corresponding welding torch) for fiducial, - 8 bit string output L place correspondence all right the least value 0, the intermediate worth that 8 bit string export picture center correspondence all right, 8 bit string output 2L length correspondence all right maximum. Computation is current groove center and standard horizontal position are poor, and translate into strings together a data, pass serial mouth output. Normal Align=center>normal >NS(j)=Tp and Ts(1:100, j:(Calculate related the each other of J+Stc)) result J=j+1j<Sbc-Stc+1? The = of center of the groove that list number of correspondence of the maximum in computational array S gets a number + (groove width / 2) Y undertakes to follow-up image small commutation reachs 2 values to change processing to obtain Ts of 2 values image; Initialization J=1normal >j, the S that list number, memory pattern plate is in sequel is when the translation in 2 values image related the each other of each corresponding position the array Tp of the value, pattern plate Ts, 2 values of follow-up groove change image Stc, the total line of pattern plate counts Sbc, 2 values image is listed always. (2) be aimed at different groove, soldering initiative condition makes pattern plate of one corresponding trends, have this pattern plate and picture of welding line groove pattern plate matchs computation, can get the basic seat of the groove in image and its center line. CNC Milling