An Improved Cascade R-CNN-Based Target Detection Algorithm for UAV Aerial Images
Hao Huang, Liangliang Li, Hongbing Ma
Abstract
Target detection in the UAV aerial images is a challenging task in computer vision. Compared with traditional images, UAV images have the characteristics of small and dense targets and complex scenes, which have been a difficult task for target detection. In this paper, an improved algorithm based on cascade R-CNN is proposed to add superclass detection on top of the original one, and then fuse the regression confidence and modify the loss function to enhance the detection capability of targets. Experiments are conducted on Visdrone dataset, and the experimental results show that our designed algorithm is simple and has good performance, which can effectively improve the detection accuracy of aerial photography targets while ensuring real-time.