In this paper, we propose mushrooms detection, localization and 3D pose estimation algorithm using RGB-D data acquired from a low-cost consumer RGB-D sensor. We use the RGB and depth information for different purposes. From RGB color, we first extract initial contour locations of the mushrooms and then provide both the initial contour locations and the original image to active contour for mushrooms segmentation. These segmented mushrooms are then used as input to a circular Hough transform for each mushroom detection including its center and radius. Once each mushroom's center position in the RGB image is known, we then use the depth information to locate it in 3D space i.e. in world coordinate system. In case of missing depth information at the detected center of each mushroom, we estimate from the nearest available depth information within the radius of each mushroom. We also estimate the 3D pose of each mushroom using a pre-prepared upright mushroom model. We use a global registration followed by local refine registration approach for this 3D pose estimation. From the estimated 3D pose, we use only the rotation part expressed in quaternion as an orientation of each mushroom. These estimated (X,Y,Z) positions, diameters and orientations of the mushrooms are used for robotic-picking applications. We carry out extensive experiments on both 3D printed and real mushrooms which show that our method has an interesting performance.
翻译:在本文中, 我们提出蘑菇检测、 本地化和 3D 构成估算算法, 使用从低成本消费的 RGB- D 传感器获得的 RGB- D 数据 。 我们使用 RGB 和深度信息用于不同的目的 。 从 RGB 颜色, 我们首先提取蘑菇的初始轮廓位置, 然后提供蘑菇切片的初始轮廓位置和原始图像, 以作为蘑菇切片的主动轮廓。 这些分块蘑菇随后被用作每次蘑菇检测的循环Hough变换的输入, 包括其中间和半径。 一旦知道每只蘑菇在 RGB 图像中的中心位置, 我们然后使用深度信息将其定位于3D 空间, 即世界协调系统中的3D 空间 。 如果每根蘑菇的检测中心缺少深度信息, 我们从每根蘑菇的半径范围内最接近的深度信息中估算出每根蘑菇的3D 。 我们还使用预制的蘑菇模型来估算每根蘑菇的3D 。 我们使用全球注册程序, 3D 来估算3D 3D 。 根据估计, 我们使用3D 配置, 我们仅使用3D 显示的旋转时, 我们仅在3D 版本中表示的轮值部分显示的试制, 方向,,, 的试根蘑菇 的试室, 直方向,,,,, 直方 。