增强边缘信息的全卷积神经网络遥感影像建筑物变化检测

doi:10.13474/j.cnki.11-2246.2023.0169

摘要/Abstract

摘要： 针对现有很多深度学习的建筑物变化检测方法未考虑图像的结构特征,导致建筑物边缘像素分割精度低的问题, 本文提出了一种增强边缘信息的遥感影像建筑物变化检测模型。首先采用Canny算法和概率霍夫变换算法提取双时相影像中建筑物的直线边缘特征图作为图像结构特征;然后将双时相影像及其对应的边缘特征图输入到增强边缘信息的全卷积神经网络(FCN)中;最后采用骰子损失和交叉熵损失加权组合函数衡量网络模型。试验表明,增强边缘信息的FCN网络在精度评价和视觉分析上具有一定的优越性。

关键词: 变化检测, 建筑物, 边缘信息提取, FCN, WHU数据集

Abstract: For many existing deep learning building change detection methods, it is difficult to obtain image structure features, which leads to the problem of low segmentation accuracy of building edge pixels. In this paper, a building change detection model based on enhanced edge information in remote sensing images is proposed. Firstly,the Canny algorithm and the probabilistic Hough transform algorithm are used to extract the linear edge feature map of the building in the bitemporal image as the image structure feature. Then the bi-temporal images and their corresponding edge feature maps are input into a fully convolutional neural network (FCN) that enhances edge information. Finally, the weighted combination function of Dice Loss and CrossEntropy Loss is used to measure the network model. Experiments show that the FCN network with enhanced edge information has certain advantages in accuracy evaluation and visual analysis.

Key words: change detection, buildings, edge information extraction, FCN, WHU dataset

中图分类号:

P237

陈婕, 刘纪平, 徐胜华. 增强边缘信息的全卷积神经网络遥感影像建筑物变化检测[J]. 测绘通报, 2023, 0(6): 61-67.

CHEN Jie, LIU Jiping, XU Shenghua. Building change detection in remote sensing images with fully convolutional neural network enhanced edge information[J]. Bulletin of Surveying and Mapping, 2023, 0(6): 61-67.

参考文献

[1] 眭海刚,冯文卿,李文卓,等. 多时相遥感影像变化检测方法综述[J]. 武汉大学学报(信息科学版),2018,43(12):1885-1898.
[2] 张良培,武辰. 多时相遥感影像变化检测的现状与展望[J]. 测绘学报,2017,46(10):1447-1459.
[3] 李德仁. 利用遥感影像进行变化检测[J]. 武汉大学学报(信息科学版),2003,28(S1):7-12.
[4] 张景发,谢礼立,陶夏新. 建筑物震害遥感图像的变化检测与震害评估[J]. 自然灾害学报,2002,11(2):59-64.
[5] 李强,张景发,龚丽霞,等. SAR图像纹理特征相关变化检测的震害建筑物提取[J]. 遥感学报,2018,22(S1):128-138.
[6] HUANG X,ZHANG L,ZHU T. Building change detection from multitemporal high-resolution Rremotely sensed images based on a morphological building index[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing,2014,7(1):105-115.
[7] 史忠奎,李培军,罗伦,等. 基于形态学属性剖面和单类随机森林分类的道路路域新增建筑物提取方法[J]. 北京大学学报(自然科学版),2018,54(1):105-114.
[8] 唐菲菲,阮志敏,张亚利,等. 基于机载LiDAR和GIS数据的建筑物变化信息自动检测方法[J]. 国土资源遥感,2016,28(1):57-62
[9] 李军胜,党建武,王阳萍. 多特征融合的高分辨率影像建筑物变化检测[J]. 测绘通报,2019(10):105-108.
[10] 吕野,胡翔云. 利用增量式马尔科夫随机场分割提取高空间分辨率遥感影像道路[J]. 国土资源遥感,2018,30(3):76-82.
[11] 冯文卿,眭海刚,涂继辉,等. 联合像素级和对象级分析的遥感影像变化检测[J]. 测绘学报,2017,46(9):1147-1155.
[12] ZHU B,GAO H,WANG X,et al. Change detection based on the combination of improved SegNet neural network and morphology[C]//Proceedings of the 3rd IEEE International Conference on Image,Vision and Computing (ICIVC).Chongqing:IEEE,2018.
[13] DAUDT R C,LE SAUX B,BOULCH A. Fully convolutional siamese networks for change detection[C]//Proceedings of the 25th IEEE International Conference on Image Processing (ICIP).Athens,Greece:IEEE,2018.
[14] JI S,SHEN Y,LU M,et al. Building instance change detection from large-scale aerial images using convolutional neural networks and simulated samples[J]. Remote Sensing,2019,11(11):1343.
[15] 顾炼,许诗起,竺乐庆. 基于FlowS-Unet的遥感图像建筑物变化检测[J]. 自动化学报,2020,46(6):1291-1300.
[16] 陈良轩,于海洋,李英成,等. 一种融合注意力机制的建筑物变化检测模型[J]. 测绘科学,2022,47(4):153-159.
[17] CANNY J. A computational approach to edge detection[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,1986,8(6):679-698.
[18] LONG J,SHELHAMER E,DARRELL T. Fully convolutional networks for semantic segmentation[C]//Proceedings of 2015 IEEE Conference on Computer Vision and Pattern Recognition(CVPR).Boston:IEEE,2015.
[19] JI S,WEI S,LU M. Fully convolutional networks for multisource building extraction from an open aerial and satellite imagery data set[J]. IEEE Transactions on Geoscience and Remote Sensing,2019,57(1):574-586.