Study On A Recogniziation Technology Of Seam Center Based On Welding Visual Image Centroid

Seam tracking in real-time welding is important to welding automatic and improved joint quality. Vision sensor is one of the most important sensors available in real-time welding seam tracking;the key to vision sensor method lies in the dynamic valid identification of seam center from welding visual image captured in welding. Methods emphasized in extraction seam center technology are filtering, binarization, edge detection and wavelet transforms as well. Image centroid is an easier variable to be measured, this dissertation intends to establish the new method of extracting seam center by building the relationship between centroid and seam center.Grey value of image is the function component of surface emissivity and the radiant intensity those are related closely to the temperature field of workpieces. Therefore, the numerical relation of image grey value and radiance of workpiece has been established based on the formula of image grey value and then, the solution between centroid and seam center has been obtained when the temperature field is available by employing numerical simulation method under a certain welding condition. Further, the inner relationship between the seam center position and image centroid value has been studied;and it is the emissivity and temperature difference between the workpiece surface and seam that make the centroid value change with seam center. The influent degree of several welding processing factors is studied by varying welding conditions and the conclusion has been arrived that the seam width, welding velocity and welding heat input have most impacts on the shape of centroid-offet model.It can be seen easily that the centroid-offset of welding visual image behavior changing with seam-offset is the most same one as sine curve after the centroid model is converted to centroid-offset one for better expression and actual practice. The centroid-offset model can be well fitted by sine curve called experiential formula with three characteristic parameters according to a number of fitted experiments. In thiscase, with those three processing factors and three characteristic parameters, spatial swatches for regressive analysis have been setup by the method of least square fitting. Further, the numerical relationship of centroid-offset value and seam-offset has been built after multi-variable regressive analysis has been carried out.Because of the complexity of welding visual method, most of all, the intrusion of intensive arc and the limitations of CCD resolution with which effect the quality of welding realtime image, further, effect the validation of experience test. Therefore, the comparative experiments is used to make sure the fittest conditions such as exposing time of CCD, selective weld pool region, filter wavelength, and so on. The centroid-offsets of welding visual images obtained in real time welding and seam-offsets are calculated and those experimental data are added to theoretical centroid-offset model. A series of comparative curve have exhibited that experimental centroid-offset data wave up and down the theoretical centroid-offset model. Same conclusion is arrived that experimental data have concorded with theoretical model very well and the characteristic of anti-intrusion has been testified at the same time.