Hough-space-based object recognition tightly coupled with path planning for robust and fast bin-picking

Abstract

The proposed bin-picking method combines object recognition with path planning. To avoid conflicts between the assumptions of the elemental techniques needed for bin-picking, object recognition is combined with path planning by using environmental information. To achieve this combination, a Hough transform, which records the model-to-image matches in a Hough space, is used to estimate the pose. The matches represent the arrangement of the objects, so they can be regarded as environmental information for path planning. To reduce the number of recognition errors and the object-detection time, a pair of object features that reduces the number of invalid votes in the Hough space is used for the Hough transform. Simulated path planning showed that using a Hough space to represent the environmental information improves the ability to plan a safe path for the manipulator. Simulated object recognition showed that using a pair of features makes the process faster and reduces the number of invalid votes. The pose estimation and safe path planning ability were confirmed by an experiment on casting objects using a range finder and a robot.