Research On Recognition Of Flange Weld Based On CCD Sensor

Now China has already become the largest international welding country. Theannual steel production has exceeded more than500million tons, half of the steel has tobe welded to get parts and products. Because of the traditional manual welding highlyneeds workers the skills of manipulation and the intensity of labor, as the development ofcomputer technology, control technology and mechanical technology, thus put forwardthe higher request of welding automation. Machine vision welding system, has thecharacteristics of high efficiency, high flexible, highly automated and can well satisfy thehigh precision and high speed of modern welding, this paper designed a set of laser visionwelding seam tracking system.Laser visual welding seam tracking system needs to carry through visual calibrationin the first place, so as to determine the geometric relationships between the worldcoordinate system and the image coordinate system. Accurate camera calibration is veryimportant for the measurement and the reconstruction precision of the stereo visionsystem. The camera calibration technique based on the perspective transformation matrix,although can meet the real-time computation of the camera, but it exists faults that is toomuch parametric. This paper, using nonlinear equation parameter estimation method onwelding model recalibration, overcomes the overmuch parameterization in the process oftraditional calibration.During the welding process, due to the strong interference from the electric arc, sothe extracted images exist noise. This paper puts forward a method of real-time imageprocessing. The method is divided into two parts: The first part is the image pre-processing algorithms, mainly includes the image add window processing, imageenhancement, smooth filtering and binarization. The second part is the image post-processing algorithm, mainly includes the extraction of the center line of the band, andthe extraction of feature points. The linear fitting for feature points prepares for thecontrol part to track welding seam.The TMS320DM642as the welding seam real-time detection system of high-speedsignal processor is established. It analyzed the working process of the system of thewelding seam image basic processing, and its hardware structure of the system.Using the robot’s internal encoder data and coordinate transformation calibrated theworkpiece coordinate system, and calculated the workpiece coordinate system relative tothe situation of homogeneous transformation matrix. Using calibration result of theworkpiece coordinate system can calculate the coordinate transformation matrix of themanipulator, through inverse kinematics of the robot six joint angles can be calculated,which can calculate the six joints pulse value needs. Using the plane line and arc interpolation principle, according to the center of the image processing algorithm fittingcan automatically track the trajectory of the welding seam.