Image Recognition of Small UAVs Based on Faster RCNN
Jingbin Zhao, Shengjun Wei, Huimin Xie, Hao Zhong · 2020
Image recognition of small UAVs is the basis of anti-UAV technology based on photoelectric detection. In this paper, an automatic image recognition method for small UAVs based on Faster RCNN is proposed. The deep residual network is adopted to extract image features of small UAVs. Then, the extracted features are input into the region proposal network, which can generate a region box that contains UAVs. The final recognition result is obtained through classification and regression. After training and testing the recognition model based on the dataset which contains thousands of images of common UAV on the market, the result shows that the recall of UAV recognition is 98%, the average precision is close to 97%, and the misdetection rate is low. The results of samples show that UAVs can be accurately recognized with less time, so the model owns good recognition performance.