基于深度学习的无人机影像建筑物自动提取

Buildings Extractied From UAV Image Based on Deep Learning

在无人机影像建筑物自动提取过程中,传统地物分类算法其精度已无法满足生产过程中的分类要求。为此,文章提出以深度学习技术结合条件随机场应用于无人机影像建筑物的自动提取方法。首先利用基于残差模块的卷积神经网络对图像进行特征提取,然后利用全卷积对图像进行反卷积,恢复图像特征。基于初步分类结果,利用条件随机场模型进行边缘细化。通过对实验结果进行分析,验证了该算法应用于无人机影像建筑物自动提取的可行性。

In the tasks of automatic building classification from UAV image,the accuracy and efficiency of traditional algorithms can’t meet the production requirements.Therefore,a method combined with the deep learning and CRF model for automatic building extraction from UAV image is proposed in this paper.Firstly,the convolutional neural network based on the residual module is used for feature extraction,as well as the full convolution operation is used for deconvolution to recover the image features.Then,the CRF model is used to refine the edges of the preliminary classification results.The results of experiments show that the proposed method is applicable for the tasks of automatic building classification from UAV image.