In the context of high-quality development

green digital technology

including artificial intelligence technology

plays an important role in promoting the development of quality inspection of power grid infrastructure. In the current power grid infrastructure construction process

the security and intelligence of quality inspection has been made realized. However

such problems such as false detection

missed detection and low detection efficiency are inevitably existed. To this end

a kind of intelligent detection method for power grid infrastructure quality based on CBAM-ASFF-YOLOv4 is proposed. Firstly

this method introduces adaptively spatial feature fusion (ASFF) into the neck network structure of the YOLOv4 algorithm to improve the original PANet and solves the problem of variable target scales by optimizing feature fusion. Then

by adding a convolutional block attention module (CBAM) after each feature layer

the feature extraction capability of the backbone network for small targets of power grid infrastructure quality detection is effectively enhanced. Finally

the dynamic weight determination method is adopted to reduce the loss ratio of accurate detection results. The simulation results of the example show that the average detection accuracy and detection speed of the quality intelligent detection method proposed in this paper reach 92.81%and 41.16 fps respectively. Compared with SSD

Faster-RCNN and YOLOv4

the detection time is shortened and the convergence speed and detection accuracy of the algorithm are improved. The method can be able to output corresponding alarm data and reports while realizing automatic identification of hidden image defects.