Visual Perception System of EROS Humanoid Robot Soccer

In this paper, a humanoid robot soccer perception system, consisting of a ball, field detection, and localization, is developed in order to deal with the new rules in RoboCup. Color segmentation and image morphology are improved together with a more sophisticated machine learning algorithm to detect a soccer ball robustly. Those algorithms are still favorable due to its real-time running in most of the embedded platform. For localization, the field is divided into pre-deðned grids and employing k-NN (k-nearest neighbor) to determine the robot location in the grids. Pre-defined grids are used to reduce computation due to matching with a map. Experiment results show that the developed system relatively well for adapting to the new rules update.

Image Morphology-Based Path Generation for High-Speed Pocketing

Pocket milling has long been a popular means for machining pocket features in structural parts and skins in the aviation industry. Recent advanced milling technologies pose new challenges for pocket milling path which existing contour-parallel path generation schemes cannot overcome. For high-speed machining, pocket milling path is desired to be smooth and with no tool retractions during the process, while the path stepover should be kept within a prescribed range to achieve relatively constant cutting load. These geometric constraints are also vital in the application of aircraft skin mirror milling in order to guarantee a correct and consistent thickness signal reception for real-time adjustment of the process. Traditional path optimization based on local modification can only meet a few of these constraints while others are being violated. Therefore, we propose a novel contour-parallel path generation scheme that respects all these process constraints by utilizing the idea of image morphology. The two-step scheme first generates an initial path by propagating from the rectified medial curve of the pocket shape. The initial path is then treated as a binary image being iteratively deformed and projected back into the pocket region via quadratic optimization. Experimental results show that our developed scheme can generate a smooth, tool retraction free and stepover-guaranteed path for various shapes of pocket.

Visual Locating of Reactor in an Industrial Environment Using the Composite Method

To achieve an automatic unloading of a reactor during the sherardizing process, it is necessary to calculate the pose and position of the reactors in an industrial environment with various amounts of luminance and floating dust. In this study, the defects of classic image processing methods and deep learning methods used for locating the reactors are first analyzed. Next, an improved You Only Look Once(YOLO) model is employed to find the region of interest of the handling hole and a handling hole corner detection method based on the image morphology and a Hough transform is presented. Finally, the position and pose of the reactors will be obtained by establishing a 3D handling hole model according to the principle of a binocular stereo system. To test the performance of the proposed method, a set of experimental systems was set up and experiments were conducted. The results indicate that the proposed location method is effective and the precision of the position recognition can be controlled to within 4.64 mm and 1.68 ° when the cameras are approximately 5 m away from the reactor, meeting the requirements.