基于超分辨率重建技术的遥感图像小目标检测

Detection for Small Target Ship in Remote Sensing Image Based on Super Resolution Reconstruction Technology

针对传统舰船遥感图像存在背景复杂和信息模糊等问题,提出一种基于超分辨率重建的舰船遥感图像小目标检测算法.该方法首先通过超分辨率重建技术对信息模糊的原始遥感图像进行清晰重建,防止重建图像过程中出现过大的固有特征损失和过于平滑的梯度变化.在此基础上建立Faster R-CNN网络,自动提取图像数据集目标特征,准确地实现舰船遥感图像中的小目标识别.实验结果表明,基于超分辨率重建的检测算法的综合效率达到65.5%,相比传统算法提高12.9%.由此证明,改进后的算法能够有效克服目标尺寸小和识别背景复杂所带来的检出率低和准确率差等问题.

Aiming at the problems of complex background and fuzzy information in traditional ship remote sensing image, a small target detection algorithm based on super-resolution reconstruction is proposed.Firstly, the original remote sensing image with fuzzy information is reconstructed by super-resolution reconstruction technology, and the excessive loss of inherent features and too smooth gradient change are prevented in the process of reconstruction.On this basis, Faster R-CNN network is established, which can automatically extract and learn the target features of training set, and accurately complete the task of small target recognition in ship remote sensing image.The experimental results show that the overall efficiency of the detection algorithm combined with super-resolution reconstruction is 65.5%,which is 12.9% higher than the traditional algorithm.Experiments show that the improved algorithm can effectively solve the problems of low detection rate and poor accuracy caused by small target size and complex background.