Brightness-Based Selection and Edge Detection-Based Enhancement Separation Algorithm for Low-Resolution Metal Transfer Images

Next-generation gas metal arc welding (GMAW) machines require the rapid metal transfer process be accurately monitored using a high-speed vision system and be feedback controlled. However, the necessity for high frame rate reduces the resolution achievable and bright welding arc makes it difficult to clearly image the metal transfer process. Processing of images for real-time monitoring of metal transfer process is thus challenging. To address this challenge, the authors analyzed the characteristics of metal transfer images in a novel modification of GMAW, referred to as double-electrode GMAW, and proposed an algorithm consisting of a system of effective steps to extract the needed droplet feedback information from high frame rate low-resolution metal transfer images. Experimental results verified the effectiveness of the proposed algorithm in automatically locating the droplet and computing the droplet size with an adequate accuracy. Note to Practitioners-Monitoring of metal transfer process is a fundamental step toward intelligent control of gas metal are welding process and its modifications. However, the metal transfer rate may exceed over 100 Hz and its monitoring requires high frame rate images so that the resolution of the image is relatively low. Due to the low-resolution and harsh welding environment, automated processing of the images for droplet identification and computation is challenging. This paper proposes a system of solutions to process the low-resolution images to obtain robust and accurate estimation of the droplet location and size. Experiments verified the effectiveness of the proposed solutions and future work will focus on algorithm optimization and high speed processor implementation to improve the speed for real-time control of droplet trajectory and size which are required for future precision manufacturing applications.