基于进化深度学习的建筑室内空间布局特征提取方法  

Feature extraction method of architectural interior space layout based on evolutionary deep learning

作　　者：吴彦 WU Yan(Department of art and design,Chu zhou Polytechnic,Chuzhou,Anhui 239000,China)

机构地区：[1]滁州职业技术学院艺术设计系,安徽滁州239000

出　　处：《湖南城市学院学报（自然科学版）》2022年第3期40-44,共5页Journal of Hunan City University:Natural Science

基　　金：滁州职业技术学院校级优秀骨干教师项目(YG2021015)。

摘　　要：建筑室内空间布局能力易受其特征影响。基于此,本文提出了基于进化深度学习的建筑室内空间布局特征提取方法,即采用边缘像素特征重组方法,确定建筑室内空间布局视觉图像点,然后建立建筑室内空间布局视觉特征参数融合模型,完成室内空间布局视觉图像预处理;通过边缘参数分布式检测方法,构建建筑室内空间布局视觉图像退化特征演化分析模型,完成建筑室内空间布局特征量的获取;设定建筑室内空间布局视觉图像稀疏性特征分解原则,确定特征模糊状态参数,完成建筑室内空间布局参数分布序列边缘参数获取;采用进化深度学习算法对参数的收敛性进行控制,实现建筑室内空间布局特征提取。结果表明,采用该方法提取建筑室内空间布局特征,精度最高约为98%,耗时约为1.1 s,相较其他方法均有提高。The ability of building indoor space layout is easily affected by the characteristics of space layout. A feature extraction method of building indoor space layout based on evolutionary depth learning is proposed. The edge pixel feature reconstruction method is used to determine the visual image points of indoor space layout, and then the fusion model of visual feature parameters of indoor space layout is established to preprocess the visual image of indoor space layout. By the distributed detection method of edge parameters,the evolution model of visual image degradation of interior space layout is constructed to obtain the feature quantity of interior space layout;the principle of feature decomposition of sparse feature of visual image of indoor space layout is set, the fuzzy state parameters of features are determined, and the edge parameters of distribution sequence of indoor spatial layout parameters are obtained;the convergence of parameters is controlled by evolutionary depth learning algorithm to realize the feature extraction of indoor space layout.The experimental results show that the highest accuracy of this method is about 98%, and the extraction time is about 1.1 s, which is improved compared with the comparison method.

关 键 词：进化深度学习 建筑 室内空间 布局 特征提取 视觉图像 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]