A binocular vision method for precise hole recognition in satellite assembly systems

The prevalent manual assembly process for satellites is marked by inefficiencies and a heightened risk of collision, thus limiting its scalability and potential for integration. To address these limitations, the adoption of industrial robots for an intelligent assembly mode has emerged as a promising alternative. Nonetheless, the implementation of an industrial robot satellite assembly system brings its own challenges, including low accuracy of the robot, large manufacturing errors of the satellite products and accumulated errors in the picking and placing process. As the main positioning feature, the detection of holes plays an important role in solving these problems. This paper introduces a novel binocular vision hole recognition system, coupled with an associated hole recognition algorithm, aiming at enhancing the detection accuracy of characteristic holes of satellite products. The proposed methodology includes the development of a confidence-based indicator to assess the quality of two-dimensional hole detection, and an iterative optimization of the fitting ellipse. Furthermore, a confidence-based neighborhood denoising two-stage fitting method is designed for accurate hole positioning. Experimental results validate the proposed methodology, showcasing a recognition accuracy of 0.060 mm and an assembly accuracy within 0.2 mm, thereby fulfilling the requirements of the industrial robot satellite assembly system.