Cascade Convolutional Neural Network Based on Transfer-Learning for Aircraft Detection on High-Resolution Remote Sensing Images

Figure 1

Architecture of two-level CCNN for large-area remote sensing rapid aircraft detection. The target detection process consists of two parts. First, the image is downsampled. The first-level CNN structure is used to classify the scene, with a sliding window on the downsampled image, block-by-block. The nontarget window image is excluded and the index position of the target window is reserved and passed to the next level of the network. The second level receives the target window and performs aircraft target detection on the original image. An RPN is used to get region proposals, and the GFC model is used to filter multiple region proposals and exclude the region proposals that do not satisfy the geometric characteristics of aircraft. Then, the Faster R-CNN classification model is used to classify the remaining region proposals to generate the target area. Finally, using overlap constraints, delete the redundant target area to get the final detection results.