Seam tracking is a key function of welding robots. At present, most welding robots work in the so-called “teach-and-playback”mode, which places great demand on the precise position and consistent shape of workpieces. They cannot be used flexibly, and cannot adapt to the ever changing working environment. Thus, it is very important to develop vision-based intelligent seam tracking systems. In this thesis, the main functions of the vision-based seam tracking systems are explored. Some relevant topics such as seam image processing, initial welding point positioning, vision model analysis, and intelligent seam tracking control are discussed. The main contributions of this thesis are as follows: Firstly, fast and effective image processing algorithms are developed according to the features of seam images. For narrow butt seam, “and”operation is used to the two subsequent images to eliminate the influence of big disturbances such as arc light, splash, and so on. According to the grey value distribution of the image, a robust center line detection method for the seam is designed. For butt seam with slope, the region of interest of the image and the adaptive threshold value are simultaneously determined by projecting the grey value of the image in two directions. For fillet seam, a fast profile following template matching method is presented. Secondly, a vision-based initial welding point positioning method is presented for narrow butt seam. At first, initialization is implemented to get the image feature of the initial welding point and the seam line slope. Then, the current seam line slope is determined by considering the image feature sample data and the position data from the encode. The reference image feature of the current initial welding point is computed based on the initialization results and current seam line slope. The control of initial welding point positioning is based on the error between the reference and the current image features. Thirdly, the image-based visual control method is adopted for seam tracking of butt seam without slope. Some functions such as image feature verification, image errors filtering and output pulses verification are added the controller to improve the reliability of the system. For butt seam with slope, the seam tracking control in horizontal direction is based on natural light based vision, and the one in vertical direction is based on laser structured light based vision. The controllers in two directions are...
修改评论