CN105522290A - Corrugated web H-shaped steel welding algorithm - Google Patents

Abstract

The invention provides a corrugated web H-shaped steel welding algorithm. The corrugated web H-shaped steel welding algorithm comprises the following steps: (S1) a camera is used for acquiring images of welded parts; and then, a canny operator is used for detecting positions of welded lines in the images; (S2) a camshift algorithm is used for tracking positions of laser points of a welding gun; and the positions of the welded lines are compared with the positions of the laser points to form complete closed-loop control so as to guarantee welding accuracy; (S3) a kalman algorithm is used for estimating next positions of the laser points of the welding gun, so that the welding gun is timely adjusted; and (S4) the positions of the welded lines are detected when the welding gun is shifted according to the estimated points; and the welding gun is timely adjusted for welding. The corrugated web H-shaped steel welding algorithm enables procedures of an H-shaped steel corrugated web production and welding process to be scientific and reasonable, improves the welding accuracy, prominently improves the product quality, and reduces the labor intensity.

Description

Ripple Sternum welding algorithm

Technical field

The present invention relates to a kind of Section Steel Production method, the process of particularly a kind of Ripple Sternum production.

Background technology

The innovative product Ripple Sternum of steel construction industry, web uses bellows-shaped instead by flat board, the local buckling that flat web produces can be avoided, thus adopt thinner steel plate to realize stronger bearing capacity, greatly reduce the steel using amount of building structure, the production capacity environment-friendly novel material of the industrial policy meeting national development " low-carbon economy ", will the existing general layout in profound influence domestic H profile steel market.Ripple Sternum product prospect market is very large, and product promotion emphasis need solve suitability for industrialized production problem.The welding procedure of current H profile steel corrugated web has various different form, and ubiquity operation is unreasonable, and production efficiency is low, and labour intensity is large, the problems such as difficult quality guarantee.

Summary of the invention

The object of the invention is to overcome the deficiencies in the prior art, provide a kind of Ripple Sternum to weld algorithm, make H profile steel corrugated web production welding technique operation scientific and reasonable, welding accuracy, product quality significantly improves, and reduces labour intensity, improves production efficiency.The technical solution used in the present invention is:

A kind of Ripple Sternum welding algorithm, comprises the steps:

Step S1, utilizes the image of camera collection welding position, then utilizes the position of sealing wire in canny operator detected image;

Step S2, utilizes the position of camshift algorithm keeps track welding gun laser spots, is made comparisons in sealing wire position and laser spot position and forms complete closed-loop control, ensures the accuracy of welding;

Step S3, utilizes kalman algorithm to estimate the next position of welding gun laser spots, welding gun is adjusted in time, improves speed and the efficiency of welding;

Step S4, according to the position detecting sealing wire while estimating a little mobile welding gun, adjustment welding gun welds in time.

The beneficial effect that the present invention is had compared with the prior art is: owing to adopting technique scheme, make H profile steel corrugated web production welding technique operation scientific and reasonable, welding accuracy, product quality significantly improves, reduce labour intensity, improve production efficiency, be applicable to the product that production multiple technologies require.

Accompanying drawing explanation

Fig. 1 is Ripple Sternum welder schematic diagram of the present invention.

Fig. 2 is algorithm flow chart of the present invention.

Detailed description of the invention

Below in conjunction with concrete drawings and Examples, the invention will be further described.

Fig. 1 is Ripple Sternum welding equipment, utilize the equipment of Fig. 1, the Ripple Sternum welding algorithm that the present invention proposes, comprises the steps: the first step, utilize camera 4 to gather the image of welding position, then utilize the position of sealing wire 2 in canny operator detected image; Second step, utilizes the position of camshift algorithm keeps track welding gun 3 laser spots, is made comparisons in sealing wire position and laser spot position and forms complete closed-loop control, ensures the accuracy of welding; 3rd step, utilize kalman algorithm to estimate the next position of welding gun 3 laser spots, welding gun is adjusted in time, improve speed and the efficiency of welding; 4th step: according to the position detecting sealing wire while estimating a little mobile welding gun, adjustment welding gun completes the welding to Ripple Sternum 1 in time.In Fig. 1, the centre of H profile steel 1 is corrugated web 101, and corrugated web 101 two ends are wing plates 102.

In the first step, utilizing the position of sealing wire in canny operator detected image, is detection sealing wire method relatively more conventional at present, not as emphasized in the application.

In second step, the position by laser spots in camshift algorithm keeps track image:

Initialize laser spots place search window, search window slightly larger than the size of laser spots, the search window larger than laser spots 20% ~ 50% usually selected; Window centroid position is:

$(u_c,v_c)=\[\frac{M_{10}}{M_{00}},\frac{M_{01}}{M_{00}}\]---\left(1\right)$

Wherein M ₀₀for the zeroth order square of search window, M ₁₀, M ₀₁first moment for search window:

$\left\{\begin{array}{c}{M}_{00}=\underset{u}{\Σ}\underset{v}{\Σ}I(u,v)\\ M_{10}=\underset{u}{\Σ}\underset{v}{\Σ}u\·I(u,v)\\ M_{01}=\underset{u}{\Σ}\underset{v}{\Σ}v\·I(u,v)\end{array}\right.---\left(2\right)$

In formula, (u, v) is the pixel coordinate in search window, and the gray value that (u, v) is corresponding in perspective view is I (u, v);

Then adjust the size of search window, the center of mobile search window is to its barycenter; Again search until operation times reach preset value or barycenter with in displacement be in the heart less than threshold value set in advance, namely think and restrained, then in next frame image, carry out new search, circulation execution is with the tracking of realization to laser spot position; The width w of new search window and length l is respectively:

$w=\sqrt{\frac{(a+c)-\sqrt{b^2+{(a-c)}^2}}{2}}---\left(3\right)$

$l=\sqrt{\frac{(a+c)+\sqrt{b^2+{(a-c)}^2}}{2}}---\left(4\right)$

Wherein:

$a=\frac{M_{20}}{M_{00}}-{u_c}^2,b=2(\frac{M_{11}}{M_{00}}-u_c{v}_c),c=\frac{M_{02}}{M_{00}}-{v_c}^2---\left(5\right)$

M ₂₀, M ₁₁, M ₀₂second moment for search window:

$\left\{\begin{array}{c}{M}_{20}=\underset{u}{\Σ}\underset{v}{\Σ}{u}^2\·I(u,v)\\ M_{11}=\underset{u}{\Σ}\underset{v}{\Σ}uv\·I(u,v)\\ M_{02}=\underset{u}{\Σ}\underset{v}{\Σ}{v}^2\·I(u,v)\end{array}\right.---\left(6\right)$

After camshift algorithm keeps track obtains the position of welding gun laser spots in present frame, the sealing wire position detected by canny operator in the position of laser spots and present frame contrasts, and according to site error adjustment welding torch position, form closed loop to realize the accurate welding of butt welded seam.

In 3rd step, in order to improve the efficiency of welding, the position of camshift being followed the tracks of the laser spots obtained inputs as the observation of kalman filtering algorithm, thus the next position of prediction laser spots, and adjust welding gun state in advance according to predicted position, thus shorten the time to welding gun adjustment in the next position welding process; Kalman filter is that one utilizes linear system state equation, by the observation data of input and output, the state of system is carried out to the filtering algorithm of optimal estimation; Two important equation of Kalman filter algorithm are respectively:

State equation:

X _t＝AX _t-1+W _t-1(7)

Observational equation:

Z _t＝HX _t+V _t(8)

Wherein: A represents systematic state transfer matrix; Z _trepresent the observation vector of t system; H and A is respectively observing matrix and the state-transition matrix of system; X _t, X _t-1the state vector of etching system when t, t-1 respectively; Process noise vector W _t-1with observation noise vector V _tbe the white noise sequence that average is 0, wherein W _tcovariance matrix Q _tfor

$Q_t=\left[\begin{array}{cccc}1& 0& 0& 0\\ 0& 1& 0& 0\\ 0& 0& 1& 0\\ 0& 0& 0& 1\end{array}\right]---\left(9\right)$

V _tcovariance matrix be R _t

$R_t=\left[\begin{array}{cc}1& 0\\ 0& 1\end{array}\right]---\left(10\right)$

Due to when carrying out target following between two frames time interval △ t shorter, can not there is larger change in target state, so the state of △ t internal object can be approximately uniform motion; Supposing the system state vector X _t=(S _x, S _v, v _x, v _v); Wherein: S _x, S _vrepresent the coordinate of target centroid in X-axis and Y-axis respectively, v _x, v _yrepresent the speed of target centroid in X-axis and Y-axis respectively;

Welding gun laser spots motion estimation process based on Kalman filter is: initialize wave filter, namely initialize position and the speed of laser spots, and the initial value of position is the position of the target that the target manually selecting to obtain or detection obtain; If speed is unknown, can 0 be set to, and records current time; Then utilize current laser spots motion state and covariance matrix, prediction welding gun laser spots position in the next frame and speed, obtain priori estimates; The final method utilizing feedback, Posterior estimator is obtained in conjunction with new actual observed value and priori estimates, and it can be used as welding gun laser spot position and the velocity information of this frame, utilize the updating location information wave filter that the tracking results of present frame camshift obtains, again predict that welding gun laser spots is in the position of next frame and speed, circulation execution is gone down.

Claims (4)

1. a Ripple Sternum welding algorithm, is characterized in that, comprise the steps:

Step S1, utilizes the image of camera collection welding position, then utilizes the position of sealing wire in canny operator detected image;

Step S2, utilizes the position of camshift algorithm keeps track welding gun laser spots, is made comparisons in sealing wire position and laser spot position and forms complete closed-loop control, ensures the accuracy of welding;

Step S3, utilizes kalman algorithm to estimate the next position of welding gun laser spots, welding gun is adjusted in time;

Step S4, according to the position detecting sealing wire while estimating a little mobile welding gun, adjustment welding gun welds in time.

2. Ripple Sternum welding algorithm as claimed in claim 1, is characterized in that:

Step S2 specifically comprises:

Initialize laser spots place search window, search window is slightly larger than the size of laser spots, and window centroid position is:

$(u_c,v_c)=\[\frac{M_{10}}{M_{00}},\frac{M_{01}}{M_{00}}\]---\left(1\right)$

Wherein M ₀₀for the zeroth order square of search window, M ₁₀, M ₀₁first moment for search window:

$\left\{\begin{array}{c}{M}_{00}=\underset{u}{\Σ}\underset{v}{\Σ}I(u,v)\\ M_{10}=\underset{u}{\Σ}\underset{v}{\Σ}u\·I(u,v)\\ M_{01}=\underset{u}{\Σ}\underset{v}{\Σ}v\·I(u,v)\end{array}\right.---\left(2\right)$

In formula, (u, v) is the pixel coordinate in search window, and the gray value that (u, v) is corresponding in perspective view is I (u, v);

Then adjust the size of search window, the center of mobile search window is to its barycenter; Again search until operation times reach preset value or barycenter with in displacement be in the heart less than threshold value set in advance, namely think and restrained, then in next frame image, carry out new search, circulation execution is with the tracking of realization to laser spot position; The width w of new search window and length l is respectively:

$w=\sqrt{\frac{(a+c)-\sqrt{b^2+{(a-c)}^2}}{2}}---\left(3\right)$

$l=\sqrt{\frac{(a+c)+\sqrt{b^2+{(a-c)}^2}}{2}}---\left(4\right)$

Wherein:

$a=\frac{M_{20}}{M_{00}}-{u_c}^2,b=2(\frac{M_{11}}{M_{00}}-u_c{v}_c),c=\frac{M_{02}}{M_{00}}-{v_c}^2---\left(5\right)$

M ₂₀, M ₁₁, M ₀₂second moment for search window:

$\left\{\begin{array}{c}{M}_{20}=\underset{u}{\Σ}\underset{v}{\Σ}{u}^2\·I(u,v)\\ M_{11}=\underset{u}{\Σ}\underset{v}{\Σ}uv\·I(u,v)\\ M_{02}=\underset{u}{\Σ}\underset{v}{\Σ}{v}^2\·I(u,v)\end{array}\right.---\left(6\right)$

3. Ripple Sternum welding algorithm as claimed in claim 2, is characterized in that:

Search window larger than laser spots 20% ~ 50%.

4. Ripple Sternum welding algorithm as claimed in claim 1, is characterized in that:

In step S3, comprise a welding gun laser spots motion estimation process based on Kalman filter:

Initialize Kalman filter, namely initialize position and the speed of laser spots, the initial value of position is the position of the target that the target manually selecting to obtain or detection obtain; If speed is unknown, can 0 be set to, and records current time; Then utilize current laser spots motion state and covariance matrix, prediction welding gun laser spots position in the next frame and speed, obtain priori estimates; The final method utilizing feedback, Posterior estimator is obtained in conjunction with new actual observed value and priori estimates, and it can be used as welding gun laser spot position and the velocity information of this frame, utilize the updating location information Kalman filter that the tracking results of present frame camshift obtains, again predict that welding gun laser spots is in the position of next frame and speed, circulation execution is gone down.