In this paper,a segmentation method based on strong semantic feature fusion is proposed for visual detection of mechanical faults in large ports.Based on DEPLABV 3+architecture,this method is innovatively improved,including dual-scale feature extraction and feature fusion attention mechanism(FFM).By using the self-collected fault data set of Guangzhou Port's large crane,the port machinery faults such as rust,paint peeling and cracks were successfully detected,showing high accuracy and robustness.The experimental results show that this method has higher accuracy and generalization ability compared with the original DeepLABV 3+architecture,which provides an effective deep learning solution for visual detection of mechanical faults in large ports,and provides new ideas and directions for future research in this field.
关键词
港口机械/故障检测/强语义特征融合/深度学习/视觉化检测
Key words
port machinery/fault detection/strong semantic feature fusion/deep learning/visual detection
维普
万方数据