高斯混合模型應用在室內設計可用性的探討

陳興忠 ； 蕭詠鍵 ； 李元榮 ； 王亭文

《NCS 2019 全國計算機會議》 2019 (2019/11) Pp. 736-740

https://doi.org/10.6927/NCS.201911.0144

摘要

一般在進行室內設計的時候，依據家具配置及特性來定義各個不同空間的擺設類型。隨著家具不同的擺設類型，透過Rhino6工具可以自動產生上百種的擺設方案。但是無論這些室內自動擺設的軟體如何便利，這些自動擺設方案有時不能夠一一符合室內設計的一般潛規則，例如:「門不能對到床」及「床頭不能開窗」等等。為了找出能分析這些潛規則的演算法，在本次研究中分別標示出以下兩種特徵：一為利用房間的床頭點與窗戶之間的角度與距離；二為床頭點與門之間角度與距離。再進一步，我們利用各種不同群集演算法去分析出這兩種特徵的分群的結果，最後再互相比較分群的結果。根據本次研究結果顯示出高斯混合模型演算法較其他群集演算法有顯著的分群效果，因此可知高斯混合模型對室內設計圖的潛規則分析是一個可行的方法。

關鍵字

群集分析 ； 高斯混合模型 ； 室內設計

並列摘要

Generally, when making interior design, furniture is placed at the various types of furnishings space according to the configuration and characteristics. With the different types of furniture, the Rhino6 tool could automatically generate hundreds of layouts. However, some of these automatic placement schemes did not meet the requirements of interior designers owing to some hidden rule, such as the door cannot face the bed, etc. In order to find an algorithm that could analyze these unspoken rules, we define two features as following: the angles and distances between both the bedside point of the room and the window; and the angle and distance between both the head and the door. According to this study, the results show that Gaussian Mixture Model is better than the other cluster algorithms. It could have effect significant clustering. Thus, we proof that Gaussian Mixture Model is a suitable analysis approach, which it could deal with the unspoken rules of interior design.

並列關鍵字

Cluster analysis ； Gaussian mixture models ； Interior design