基于CNN模型的高分辨率遥感图像目标识别

遥感图像目标识别作为当前遥感图像应用领域中的主要研究内容，具有重要的理论意义和广泛的应用价值。近年来，深度学习成为机器学习领域的一个新兴研究方向，卷积神经网络(convolutional neural networks,CNN)是一种得到广泛研究与应用的深度学习模型。提出一种基于CNN模型的光学遥感图像目标识别方法，在传统LeNet-5网络结构的基础上，引入ReLU激活函数代替传统的Sigmoid函数和tanh函数，使用卷积展开技术将卷积运算转换为矩阵乘法，并对网络结构进行调整优化，提高目标识别的准确性和效率。利用Quick Bird上的0.6 m分辨率的遥感图像进行验证，实验结果表明，基于改进的CNN模型的方法可以取得较高的目标识别准确率和效率。

Remote sensing image target recognition based on CNN

Remote sensing image target recognition is of greatly theoretical significance and wide application value. In recent years, Deep Learning has become a new research direction in the field of machine learning, and Convolutional Neural Network (CNN) is a deep learning model of widely research and application. In this paper, a method of optical remote sensing image target recognition based on CNN model is proposed. Based on the traditional LeNet-5 network, ReLU activation function is used to replace the Sigmoid and tanh activation functions. Using convolutional layer unrolling technique, which makes the convolutional calculation be represented by matrix multiplication. The parameters and the structure of the CNN network are adjusted to make it optimal. The accuracy and effectiveness of the proposed method are verified by using Quick Bird 0.6m resolution on remote sensing image.