Object Detection in Aerial Images Based on Cascaded CNN

Object detection in aerial images is widely used for military applications, such as reconnaissance, target surveillance, battle damage assessment, et al. However, the tasks are very challenging due to a lot of factors, such as illumination variance, scene complexity, and platform motion. To deal with these problems, a new cascaded convolutional neural network (CNN) model for object detection from airborne videos is proposed. The proposed framework adopts a cascaded structure with three levels of deep CNNs that predict objects in a coarse-to-fine manner. The experimental results showed that the proposed method can achieve better performance.