基于示例学习的铸件自动浇注系统  

Automatic Pouring System for Metal Casting based on Learning from Examples

作　　者：王永国[1] 苗立刚[2] 

机构地区：[1]安徽大学数学科学学院,安徽合肥230039 [2]东北大学秦皇岛分校电子信息系,河北秦皇岛066004

出　　处：《铸造技术》2010年第11期1503-1506,共4页Foundry Technology

基　　金：安徽大学2112项目资助(KJTD002B)

摘　　要：采用PTZ摄像机监控金属铸件的浇注生产线,并提出了一种基于示例学习的自动浇注算法,它包括示例训练和自动浇注两个阶段。训练阶段由人工操作属铸件的浇注过程作为示例,利用多帧视频图像的累加来估计浇注位置、浇口杯面积、浇注时间等参数。在自动浇注阶段,对每帧图像采用投影分析来消除液柱的干扰,并综合利用浇注面积和浇注时间对塞棒的开闭进行控制,同时根据浇注参数的变化调整PTZ摄像机的运动,使得浇注过程始终位于摄像机的视场中心附近。实验表明,该自动浇注系统具有较高的稳定性和精度,能够有效提高浇注生产线的生产效率。PTZ Camera is used to acquire image sequences of metal cast pouring process,and an automatic pouring algorithm is proposed based on learning from examples. This algorithm includes two stages： training from examples and automatic pouring. At the first stage,it takes manual operation of metal cast pouring as examples,and takes multiple frames to estimate pouring position,pouring basin area,and pouring time. In the second stage,it combines pouring basin area and pouring time to control the open/close of sprue plug,and it updates pouring parameters to direct the motion of PTZ camera automatically,thus,to make pouring process lie around the center of PTZ camera views all the time. Experiment results show that this system achieves high robustness and precision,and improves the efficiency of product line of metal casing greatly.

关 键 词：自动浇注 示例学习 铸件浇注 图像分割 

分 类 号：TG233[金属学及工艺—铸造]