论文部分内容阅读
提出一种CNN的遥感图像飞机检测的方法。首先获得预训练好的CNN,然后通过参数迁移获得五层卷积层模型参数,接着利用遥感图像对第五层卷积层进行微调获得一个特征提取器。将特征提取器用于提取遥感图像训练集的深度特征,训练可变形部件检测模型。实验表明,提出的方法大大提高了遥感图像飞机目标检测精度,准确率达96%以上。
A CNN remote sensing image plane detection method is proposed. Firstly, a pre-trained CNN is obtained, and then five-layer convolutional model parameters are obtained through parameter migration. Then, a remote-sensing image is used to fine tune the fifth layer convolution layer to obtain a feature extractor. The feature extractor is used to extract the depth features of the remote sensing image training set and train the deformable component detection model. Experiments show that the proposed method greatly improves the accuracy of target detection in remote sensing images with an accuracy of over 96%.