Study On A New Approach Of Seam Detection Based On Vision Sensor In Automatic Welding Process

Seam tracking control is the key during welding process, which plays an important role in the automatic control of welding quality. And seam detection is one of the kernel technologies of seam tracking. Many types of sensors were applied to seam detection process, which including vision sensor, arc sensor, infraded sensor and so on. It is a trend technology that implementing seam detection using vision sensor.In traditional seam tracking system based on vision sensors, the weld seam located in the front of weld pool was captured as characteristic information of weld seam. And the errors between the seam center and the arc was gained by image processing. Unlike traditional method which simply analyze the edge of weld seam, a new approach for seam tracking is put forward in this dissertation, which processes the whole region of weld pool image and extract the centroid coordinates as the variable. A model to determine seam position according to the centroid position was establish through analyzing the characteristic of centroid position, distribution of pixel gray and threshold.An experimental device was designed and established to detect seam automatically employing vision sensor, stepper motor driver and the industrial PC control system. In this dissertation, the edge detection operators including Roberts, Sobel, Kirsch, Prewitt, Guass-Laplace and Canny were employed to process the weld seam image. The results showed that Canny operator is suitable for seam tracking process. Finally, Canny edge detection operator was applied to seam detection, and the original seam curve was obtained which provide value of centroid position.In order to study the influence of distribution of pixel gray and threshold on centroid position, single snap weld experiments were conducted in different technical parameters. These experiments mainly analyze the relation of them and got an...