摘要
针对传统机器学习方法提取建筑物,耗时长和精度低的问题。文中选用深度学习中的SegNet语义分割模型进行算法改进,提出了一种基于稀疏约束SegNet的高分辨率遥感影像建筑物提取算法。首先对SegNet模型加入正则项和Dropout,大大降低了模型过拟合现象的发生;其次为了模型能够提取更丰富的语义特征,算法引入金字塔池化模块;最后对SPNet模型引入Lorentz函数稀疏约束因子,构造新的语义分割模型LSPNet.为了验证提出算法的可靠性和适用性,使用优化LSPNet模型对高分辨率数据集中的建筑物识别和提取。实验结果表明,该方法与传统机器学习方法相比较,有着快速收敛、精度高的优势,并且具有很好的应用前景。
In view of the problem of time-consuming and low precision in extracting buildings caused by using traditional machine learning methods,this paper selects the SegNet semantic segmentation model in deep learning for algorithm improvement.A high-resolution remote sensing image building extraction algorithm based on sparse constraint SegNet is proposed.First,regularization and dropout are added to the SegNet model,which greatly reduces the occurrence of overfitting of the model.Secondly,in order for the model to extract richer semantic features,the algorithm introduces the pyramid pooling module.Finally,the Lorentz function sparse constraint factor is introduced to the SPNet model to construct a new semantic segmentation model LSPNet.In order to verify the reliability and applicability of the proposed algorithm,the optimized LSPNet model is used to identify and extract buildings in high-resolution data sets.Experimental results show that compared with the traditional machine learning method,the method in this paper has the advantages of fast convergence,high accuracy,and good application prospects.
作者
张春森
葛英伟
蒋萧
ZHANG Chun-sen;GE Ying-wei;JIANG Xiao(College of Geomatics,Xi’an University of Science and Technology,Xi’an 710054,China)
出处
《西安科技大学学报》
CAS
北大核心
2020年第3期441-448,共8页
Journal of Xi’an University of Science and Technology
基金
陕西省自然科学基金(2018JM5103)。
关键词
深度学习
特征提取
语义分割
稀疏约束
deep learning
feature extraction
semantic segmentation
sparse constraint
