A real-time and robust feature detection method using hierarchical strategy and modified Kalman filter for thick plate seam tracking

Abstract

Feature detection is an essential and important part in weld seam tracking of automated welding robots. In thick plate seam tracking, a profilometer based on structured light is commonly employed. Features from light stripe of the structured light will be extracted and used as primary information in visual servoing control. The accuracy, robustness, and computational cost are the main aspects of the feature detection. They directly affect the quality of the tracking. In this paper, cross mark created by cross line structured light (CLSL) is taken into account, and handled as the feature. It can be used as a pinpoint for seam tracking. Firstly, the cross mark is hierarchically estimated by coarse-to-fine strategy. In coarse estimation step, the random sample consensus (RANSAC) algorithm is applied to compute the feature position. Subsequently, the mean shift algorithm is used to estimate the precise feature position in the fine estimation step. Finally, the robustness of the detection is improved by the modified Kalman filter algorithm. The experimental results verify that the feature position estimated by the proposed method is robust. Moreover, the coarse-to-fine strategy can reduce a huge computational cost in the detection. And therefore, the detection method is proper for being used in real-time thick plate seam tracking.