论文部分内容阅读
提出了一种基于深度神经网络的机械臂最优抓取位置检测方法.相比传统手工设定的特征,基于深度神经网络的方法学习得到的特征具有较强的鲁棒性和稳定性,能够适应训练集中未曾出现的新物体.本方法首先使用基于深度学习的目标检测算法对图像中的目标物体进行检测,记录目标的类别和位置.然后根据分类检测结果,使用基于深度学习的机械臂抓取方法进行抓取位置学习.仿真实验表明所提方法能对图像中的目标物体进行较为准确的分类,在Universal Robot 5机械臂上得到的抓取实验结果证明了所提方法的有效性.
A method based on deep neural network to detect the optimal position of robotic arm is proposed.Compared with the traditional hand-set features, the feature learned by the method based on deep neural network has strong robustness and stability, In order to adapt to the new objects which do not appear in the training set, this method first uses the target detection algorithm based on depth learning to detect the target object in the image and record the target category and location.According to the classification test results, And the method is used to study the position of the grasping.The simulation results show that the proposed method can classify the target objects in the image more accurately and the results of the grasping experiment on the Universal Robot 5 robot prove the effectiveness of the proposed method.