基于卷积神经网络的摄像机姿态感知系统设计

该文设计一种基于卷积神经网络的摄像机姿态感知系统,运用深度学习方法结合传感器技术,获取摄像机实时姿态数据,特别是摄像机运动过程中的姿态数据。系统采用孪生卷积神经网络,通过摄像机采集的环境图像对孪生卷积神经网络进行训练获得摄像机姿态感知模型,在使用时通过将摄像机采集的视频图像输入摄像机姿态感知模型获得摄像机的位姿数据。系统解决可转动式摄像机的实时姿态感知问题,可在公共安全、工厂、交通和矿山等领域广泛推广应用。

This paper designs a camera posture perception system based on convolution neural network.Using deep learning method and sensor technology,real-time camera posture data,especially during camera motion,is obtained.The system uses Siamese convolution neural network to train the twin convolution neural network through the environment image collected by the camera to get the camera posture perception model.When using the system,the camera posture data is obtained by inputting the video image collected by the camera into the camera posture perception model.The system solves the real-time posture perception problem of rotatable cameras and can be widely used in public safety,factories,transportation,mining and other fields.