부산대학교 도서관

Aiming at the high rate of missed detection and false detection rate of small targets in UAV aerial photography in target detection, this paper proposes a small target detection method based on improved YOLOv3.Based on the YOLOv3 model, the 8-fold downsampled feature map output from the network is stitched with the 4-fold downsampled feature map to create a new $104 \times104$ scale detection layer. A new feature fusion network, BiFPN, is introduced to enhance feature extraction. The proposed algorithm is tested in simulation experiments on VisDrone2019 dataset, and the experimental results show that the model improves 7% over the base model with almost no impact on speed, and the detection accuracy of small targets is significantly improved.