A Detection Method of Bolt Out-of-Stock Defect in Distribution Network Lines Based on Attention Mechanism

A FCOS detection model for bolt out-of-stock defect detection of distribution network lines with attention mechanism is proposed. The spatial adaptive and size adaptive attention mechanism is mainly used to improve the Head of the original model. While improving the detection accuracy of the model, it also effectively reduces the memory usage and detection time of the model, so that the model can run on the UAV platform and detect the bolt out-of-stock defect target on the distribution network lines efficiently and accurately. Through experimental verification, the method proposed in this paper improves the accuracy of model detection under the premise of effectively reducing the calculation amount of the algorithm.