摘要
矿山检修区工作人员对矿车等机电设备的维修涉及大量的近距离手工操作,具有一定的危险性,安全巡视人员手动控制云台摄像机追踪观察检修车辆存在很大的不便。针对这一情况,采用基于深度强化学习的云台自动追踪检修区矿车方法以辅助安全巡视人员。该方法模拟了矿车在视频画面中的目标框伴随云台相机的移动而变化的虚拟环境,使用近端策略优化(PPO)算法训练强化学习模型控制云台相机转动。经过实验验证,该算法可自动控制云台相机追踪检修区矿车,使其位于画面中的合适位置,对复杂场景下的云台自动追踪具有一定的普适性。
The maintenance of mechanical and electrical equipment such as trucks by the staff in the mine maintenance area involves a large number of close manual operations,which is dangerous.It is very inconvenient for the safety inspector to manually control the PTZ camera to track and observe the maintenance vehicles.In view of this situation,a method of automatic tracking of trucks in the maintenance area based on deep reinforcement learning is proposed to assist safety inspectors.This method simulates the virtual environment in which the target frame of the truck in the video picture changes with the movement of the PTZ camera,and uses the proximal policy optimization(PPO)algorithm to train the action decision network to control the rotation of the PTZ camera.After experimental verification,the algorithm can automatically move the pan tilt to track the truck in the maintenance area so that it is in the right position in the picture,and has certain universality for PTZ automatic tracking in complex scenes.
出处
《工业控制计算机》
2023年第12期78-80,共3页
Industrial Control Computer
