Project description:针对铸坯低倍缺陷样本少、缺陷长宽比大、同类缺陷等级差异难以定量描述等挑战，开发以深度学习方法为主，融合小样本对比学习、视觉注意力机制和人工特征等关键技术，同时开发连铸坯坯号识别模型、图像在线传输子系统等实现从委托获取、图像传输、模型调用、评级、审核和报表查询等铸坯低倍缺陷智能检测系统
Classification of Disciplines:Engineering
Wanghongbing
+
Professional Title:Professor
Supervisor of Master's Candidates
Supervisor of Doctorate Candidates
Research Projects
铸坯低倍缺陷智能检测系统