An Optimal Image Processing Method for Simultaneous Detection of Weld Seam Position and Weld Gap in Robotic Arc Welding

1. Introduction

Nowadays robots have been utilized for wide range of application in manufacturing industries like transportation, pick and place or manipulation etc. which basically leads to automate the system (Deepak, Parhi, & Raju, 2014; Deepak & Parhi, 2013; Deepak & Parhi, 2016). The use of robot welding operation especially industries involving assembly lines, dangerous work environment and difficult to reach position leads to automation of welding processes. The task of automating the welding operation requires sensor integration, coordination with the welding power source and motion control. In robotic welding, different sensors are utilized for automatic robot path planning for obtaining quality welds. Vision sensors have been utilized by different researchers and industries for obtaining weld seam path and also for obtaining weld seam dimensions for weld parameter control during the time of welding. For this the images of weld seam captured by vision sensors needs to be processed and analyzed in detail. Ge et al. (2005) proposed an image processing technique consisting of diagonal differencing operator for removal of noises, binarizing, and extraction of edges to find the characteristic points of weld seam for steel pipe welding. Nele et al. (2013) developed image acquisition process for automatic finding of weld seam positions. For this template matching method has been utilized for auto finding of weld start positions. Zhou et al. (2006) proposed an image processing method consisting of steps like median filtering, thresholding, thinning and feature point extraction. Similarly, an image processing algorithm has been proposed for extraction of weld seam features from the weld image following the steps like: median filtering, thresholding, Roberts’s edge detection technique, thinning and non-linear square method (Shen, Lin, Chen, & Li, 2010; Shen, Lin, & Chen, 2007). Shi eta al (2007) utilized Canny operator for weld seam edge detection and further proposed an algorithm for weld seam extraction by using shift window. Chen et al. (2011) used threshold segmentation algorithm, edge detection and curve fitting for image processing to obtain weld seam features. A peak current self-adaptive regulating controller has been designed for weld parameter control during the time of welding. Kuo and Wu (2002) combined Prewitt operator with threshold method for edge detection and further applied Fuzzy method for selection of weld seam. Gao and Na (2005) proposed Kalman filtering based on centroid of weld pool image for extraction weld seam position. Dinham and Fan (2013; 2014) developed an image processing algorithm for auto-identification and extraction of weld seam characteristics. An image matching and triangulation by using two-dimensional homography method have been used for finding the results. They further developed a line growing algorithm for weld seam identification without any prior knowledge about weld seam.