摘要
该款仿生机械臂基于Google人工智能系统TensorFlow的深度学习开发平台,采用Inception_v3模型,并结合神经网络算法与PID控制理论技术对物体进行更高精度的识别与抓取。平台通过摄像头读取物体运动状态目标点,由神经网络生成期望轨迹,并添加递归神经网络优化抓取行为,可精准识别人体手势运动的情况并得出手势运动数据,在搭建通信网络,将手势运动数据远程传输到机械臂控制器,结合机械臂运动控制算法,使机械臂完成相应的人体手势仿生动作。实验结果表明,训练好的模型在识别手势图像时具有非常好的鲁棒性。仿生机械臂具有智能化、精度高、便于操作的优势,可代替人类来完成特定工作,更能满足信息化时代发展的需要。
This bionic manipulator is based on the deep learning development platform of tensorflow,an artificial intelligence system of Google.It adopts the concept﹣V3 model,and combines the neural network algorithm and PID control theory technology to identify and grasp objects with higher accuracy.The platform reads the object motion state target point through the camera,generates the desired trajectory by the neural network,and adds the recurrent neural network to optimize the grabbing behavior,which can accurately identify the human gesture movement and obtain the gesture movement data.In the process of building the communication network,the gesture movement data is transmitted to the manipulator controller remotely,and the manipulator is completed by combining the manipulator movement control algorithm Corresponding human gesture bionic action.The experimental results show that the trained model is very robust in the recognition of gesture images.Bionic manipulator has the advantages of intelligence,high precision and easy operation,which can replace human beings to complete specific work and meet the needs of the development of information age.
作者
陈苏明
高正创
王若愚
卢小康
叶子夜
Chen Suming;Gao Zhengchuang;Wan Ruoyu;Lu Xiaokang;Ye Ziye(Xu Hai College,China University of Mining and Technology,Xuzhou Jiangsu 221008,China)
出处
《信息与电脑》
2020年第5期116-118,共3页
Information & Computer
基金
江苏省大学生创新创业项目项目名称“基于图像识别的仿生机械臂”(项目编号:201913579008Y)。
关键词
图像识别
卷积神经网络
机械臂
仿生
image recognition
convolutional neural network
manipulator
bionics
