Multi-class Object Detection in Fusion Remote Sensing Images Based on Deep Learning
Tao Sun, Kun Liu · 2021
Deep learning has been applied successfully in the machine visual field, and it has already surpassed humans in object detection of images. Object detection is a significant content of remote sensing image analysis and is a key link to transform image data into application results. Based on a self-made high-resolution remote sensing image dataset, the Faster RCNN algorithm is used in the two-stage method and the RetinaNet algorithm in the one-stage method to carry out experiments. For the small and dense object in the remote sensing dataset, three improvements are proposed, and finally an object detection model is proposed with strong generalization ability. The results show the proposed model proposed can achieve better results in detection accuracy and computational overhead.