乡村景观改造高度集聚特征提取算法  

Highly Clustered Feature Extraction Algorithm for Rural Landscape Renovation

作　　者：杨志勇[1] 郭宗平[2] YANG Zhi-yong;GUO Zong-ping(Taiyuan Normal University,Jinzhong Shanxi 030619,China;School of Architecture,Taiyuan University of Technology,Taiyuan Shanxi 030002,China)

机构地区：[1]太原师范学院,山西晋中030619 [2]太原理工大学建筑学院,山西太原030002

出　　处：《计算机仿真》2025年第3期145-149,共5页Computer Simulation

基　　金：山西省哲学社会科学基金项目(2019B356);山西省社科联2023-2024年度重点课题研究项目(SSKLZDKT2023113)。

摘　　要：在乡村景观改造中,空间数据往往会受到噪声的影响,导致高度集聚特征的提取受到干扰或误差,使得对景观改造高度集聚特征提取精度降低。为了准确提取景观改造高度集聚特征,提出一种乡村景观改造高度集聚特征提取算法。使用遥感数据采集景观改造图像,通过非下采样Contourlet变换(Nonsubsampled contourlet transform,NSCT)提取图像不同方向的NSCT域系数,并通过改进BayesShrink阈值处理方法以去除噪声。应用模糊理论构建模糊阈值函数,处理NSCT域系数,得到去噪后的景观改造图像。引入局部二值模式算法(Local Binary Patterns,LBP)生成景观改造低密度特征图,利用旋转不变原则对特征图展开转换。通过滑动窗口遍历每个对象,统计各个对象在不同模式下的量级,得到LBP特征。将不同特征组成多维特征向量,并输入到支持向量机以提取景观改造的高度集聚特征。实验结果表明,所提方法可以得到高精度和高效率的乡村景观改造高度集聚特征提取结果。In rural landscape renovation,spatial data is often affected by noise,leading to interference or errors in the extraction of highly aggregated features,resulting in a decrease in the accuracy of extracting highly aggregated features for landscape renovation.In order to accurately extract the highly aggregated features of landscape renovation,a rural landscape renovation highly aggregated feature extraction algorithm was proposed.Firstly,remote sensing data were used to collect the images of landscape reconstruction.Then,nonsubsampled contourlet transform(NSCT)was used to extract NSCT domain coefficients in different directions.At the same time,improved BayesShrink threshold method was adopted to remove noise.Next,a fuzzy theory was adopted to construct a fuzzy threshold function and process the NSCT domain coefficients,thus obtaining a denoised image.Furthermore,Local Binary Patterns(LBP)algorithm was introduced to generate a low-density feature map of landscape reconstruction.Based on the principle of rotation invariance,the feature map was transferred.After that,the sliding window was used to traverse each object,and the magnitude of each object under different patterns was calculated to obtain LBP features.Finally,different features were composed into a multi-dimensional feature vector and inputted into a support vector machine to extract the highly clustered features of landscape reconstruction.The experimental results show that the proposed method can achieve highly clustered feature extraction with high precision and efficiency.

关 键 词：乡村景观改造 高度集聚特征 多维特征向量 局部二值模式算法 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]