景观空间梯度变化特征的提取与仿真  

Extraction and Simulation of Landscape Spatial Gradient Change Features

作　　者：刘娟[1] 张圆圆[1] 李腾 王崇宇 LIU Juan;ZHANG Yuan-yuan;LI Teng;WANG Chong-yu(College of Architecture and Design,Hebei Polytechnic Institute,Hebei Shijiazhuang 050000,China;Urban and Rural Construction Institute,Hebei Agricultural University,Hebei Baoding 071001,China)

机构地区：[1]河北工程技术学院建筑与设计学院,河北石家庄050000 [2]河北农业大学城乡建设学院,河北保定071001

出　　处：《计算机仿真》2022年第5期212-215,379,共5页Computer Simulation

摘　　要：采用目前方法提取景观空间的梯度变化特征时,无法消除噪声对特征提取产生的影响,导致方法存在抗畸变能力差、适应性差和提取精度低的问题。提出考虑异质性的景观空间梯度变化特征提取方法,对景观空间的异质性进行考虑,结合稀疏表示方法和边缘模板匹配方法获取景观空间图像,将香农信息量理论引入主成分分析方法中,对获取的景观空间图像进行降维处理。采用Haar小波变换方法对景观空间图像进行分解处理,获得高频小波稀疏和低频小波稀疏,将模糊逻辑方法引入Haar小波变换过程中,根据小波系数建立隶属度函数,计算模糊区域在模糊特征空间中的激活强度值,归一化处理激活强度值,获得景观空间的梯度变化特征。仿真结果表明,所提方法的抗畸变能力好、适应性高、提取精度高。Recently,there are many disadvantages in extracting the gradient change features of landscape space,such as anti-distortion performance,adaptability and low extraction accuracy caused by unable to eliminate noise.The feature extraction method of landscape spatial gradient change based on heterogeneity was reported.The heterogeneity of landscape space was introduced.The sparse representation method and edge template matching method were adopted to obtain landscape spatial images.Shannon’s information theory was introduced into the principal component analysis method for reducing the dimension of landscape spatial images.According to the Haar wavelet transform method,the landscape spatial image was segmented to obtain high-frequency wavelet sparse and low-frequency wavelet sparse.Haar wavelet transform and fuzzy logic method were combined.Based on the wavelet coefficients,the membership function was founded to calculate the activation intensity value of the fuzzy region in the fuzzy feature space,and normalize the activation intensity value.Eventually,the gradient change feature of the landscape space was obtained.The simulation results show that this method has good anti-distortion performance,excellent adaptability and high extraction accuracy.

关 键 词：异质性 景观空间 香农信息量 降维处理 特征提取 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]