摘要
针对现有输送带检测方法无法适应小样本条件和复杂工业环境导致精度差的问题,提出一种更可靠的输送带表面损伤检测方法ACGAN-BELT。基于ACGAN网络实现输送带表面损伤样本的多类别增广,构建基于YOLOX的检测网络对输送带损伤信息进行细粒度特征感知。实验表明,ACGAN-BELT方法在原始工况和复杂工况2种工况下具备更好的可靠性和泛化性。
Aiming at the problem of poor accuracy caused by the inability of existing conveyor belt detection methods to adapt to small sample conditions and complex industrial environments,a more reliable conveyor belt surface damage detection method ACGAN-BELT is proposed.Implement multi class augmentation of surface damage samples on conveyor belts based on ACGAN network,and construct a YOLOX based detection network for fine-grained feature perception of conveyor belt damage information.The experiment shows that the ACGAN-BELT method has better reliability and generalization under original and complex two operating conditions.
作者
杨仲
杨泽坤
王贡献
Yang Zhong;Yang Zekun;Wang Gongxian(Wuhan K-crane Ocean Lifting Technology Co.,Ltd.;School of Transportation and Logistics Engineering,Wuhan University of Technology)
出处
《港口装卸》
2024年第5期1-3,12,共4页
Port Operation
关键词
输送带
损伤检测
生成对抗网络
数据扩充
conveyor belt
damage detection
generative adversarial network
data expansion
