| As a major processes in manufacturing,welding has made a great contribution.In the traditional manual welding process,the bad welding environment will cause health damage.Therefore,welding automation is essential.To solve the problems of low accuracy and slow speed of weld seam detection and location.A monocular vision system based on laser light source is designed.Combining saliency target detection and image recognition algorithms,a weld seam detection and localization method is proposed.The specific research is as follows.1)For the problem that the welding robots need to locate the initial point artificially before welding.A template matching algorithm with fused wavelet pyramid is proposed,to search for the most similar sub-image to the template image as the initial position.Through experimental analysis,the algorithm can improve the efficiency while ensuring the matching accuracy.2)For the problem that the laser band is too wide,a laser extraction method based visual attention mechanism is proposed.First,the color and the direction feature map of laser image are extracted,and then fuse two feature maps by multi-layer cellular automata to get the salient area.Through experimental comparison,the proposed algorithm can better extract the complete laser welding stripe.3)For the problem of time-consuming and low accuracy of welding location,a method combining template matching and corner detection is proposed.First,the template matching is improved for the initial localization.Extract the sub-pixel corner points of the refined weld image to get the precise localization.Through experimental comparison,the algorithm can improve the accuracy while increasing the efficiency.In summary,this paper effectively extracts the laser welding and locates the feature points through the laser vision system,which improves the current problems of low accuracy and long time in automated welding and has good application prospects.Figure 49;Table 6;Reference 49... |
PDF Full Text Request