Research Of Wield Track Control System Based On Machine Vision

This dissertation studies the arc-weld tracking system, which makes preparations for the automatic welding robot for Guangzhou TV tower. The system is composed of three parts, image acquisition subsystem, image process subsystem and motion controlling subsystem.Using structure laser as the active light source can improve SNR (Signal Noise Ratio) of image acquisition effectively. After comparing the effect and velocity, the dissertation chooses Gauss filter to weaken the noise. In addition, a narrowband filter, which is placed in front of the camera, can weaken the arc light while preserving the laser stripe effectively. The system uses color camera and color image board to acquire image. In the image pre-process stage, the pre-process module only abstracts the red quantum. Through that step, specific wavelengths of light is strengthened with others been weakened.The image process has three parts: threshold transform, edge detection and characteristic quantity recognition. After the threshold transform, the weak noise is eliminated while the laser stripe is reserved. After comparing some algorithm of edge detecting and testing their velocities, the dissertation chooses Mathematical morphology. When recognizing characteristic quantity, it can enhance fault-tolerant by black pixel accumulation algorithm.Motion card is used as the main part of motion controlling subsystem to save development time, while satisfies the accuracy. The system chooses the linear interpolation mode and it can easily track the seam with all the position known. The industrial PC is the control center of the arc track controlling system, which gets position by image process and control the welding torch position.