Communication Base-Station Antenna Detection Algorithm Based on YOLOv3-Darknet Network

Aiming at improving the efficiency and the accuracy of the base station antenna inspection, a new method is proposed to use the deep learning algorithm YOLO v3 algorithm combined with geometric mathematics to detect the downtilt angle of the antenna in our paper. The method uses Darknet-53 network to extract antenna features and multi-scale border prediction. The feature extraction is based on Darknet-53 network, and the network is inspired by the residual network. The method proposed can detect target in real time and has obvious advantages in speed and accuracy compared to Fast-RCNN. The method proposed also has weak point in the detection of antenna.