Research on Bolt Defect Detection Based on Dual-attention Mechanism and Cascade Detection Framework
The traditional bolt defect detection on transmission lines suffers from low accuracy and ef-ficiency,and this paper proposes a bolt defect detection model based on dual-attention mechanism and cas-cade detection framework.Firstly,a cascade framework for bolt defect detection is constructed and the dataset is expanded by an image augmentation method based on the GridMask;secondly,in order to im-prove the accuracy of bolt defect detection,a knowledge-guided detection model for power line fittings is proposed,the detection of the first level of gold fixtures is realized through the construction of implicit and display modules;then,the results of the gold fixture detection are inputted into the defect detection net-work for bolts at the second level,while a bolt defect detection model based on dual-attention mechanism is proposed for bolt defect detection,through the design of the multi-scale attention module,the spatial at-tention map is fused to realize the feature enhancement on the global field of view of the bolt and improve the accuracy of detection;finally,it is validated by experiments,which show that the average accuracy of bolt defect detection of the model proposed in this paper is high and the detection speed is fast.The model has good practicality,robustnessandcertain application prospects.
国家科技期刊平台
NSTL
万方数据
维普
