融合多特徵與個人化模組之畫作推薦系統;Art work recommendation system with fusion of multiple features and personalized modules

NCU Institutional Repository > 資訊電機學院 > 資訊工程研究所 > 博碩士論文 > Item 987654321/74681

jsp.display-item.identifier=請使用永久網址來引用或連結此文件: http://ir.lib.ncu.edu.tw/handle/987654321/74681

题名:	融合多特徵與個人化模組之畫作推薦系統;Art work recommendation system with fusion of multiple features and personalized modules
作者:	廖育萱;Liao, Yu-Syuan
贡献者:	資訊工程學系
关键词:	畫作風格推薦;Sobel邊緣偵測;Canny edge邊緣偵測;二維離散小波轉換;人臉偵測
日期:	2017-07-24
上传时间:	2017-10-27 14:36:04 (UTC+8)
出版者:	國立中央大學
摘要:	科技日新月異的發展，影像處理 (Image Processing)被廣泛的應用在各個領被廣泛的應用在各個領域中，車輛偵測、圖形識別人臉辨等。近日來利用空閒時欣賞畫作的們亦日漸趨多，但目前較少能直接使用圖片影像為輸入進行畫作相似風格檢索我們提供一個應用程式，藉由擷取畫作的特徵再透過相關性排序(Rank)後，即可推播出資料庫中與詢問的畫作相似風格給予使用者。後，即可推播出資料庫中與詢問的畫作相似風格給予使用者。本論文使用紋理 (Texture)、顏色 (Color histogram)、Canny邊緣偵測 (Canny edge detection)、Sobel邊緣偵測的梯度強分佈特性 (Sobel gradient distribution magnitude)等方式作為畫作的特徵。由於每個人對畫作的風格為主觀意識，因此自行設計出一個評價應用程式 (Feedback application)，經由此應用程式可獲得回饋資料作為個人化模組 (Personalized module)的訓練資料，經由畫作特徵的;The number of people visiting the art exhibition at their free time has been increased in recent years. People might be interested of a certain artwork and would like to know if there are any other artworks with similar styles. However, artwork recommendation systems that can retrieve images with similar styles are rare in the existing research works. We propose an application that can utilize the features extracted from the art works and rank the artworks by the features of the relevance and then return artworks with similar styles in our database with the query image of art work to the users. The proposed system utilizes gradient distribution, colors, texture, and face information as the features of the art works. Everyone has a personal opinion about the artworks, so we also propose a feedback mechanism. We can obtain the feedback data and the feedback data as the training data of personalized modules and train the weights of the features. Afterwards, the system can recommend artworks according to personalized preferred similar styles.
显示于类别:	[資訊工程研究所] 博碩士論文

文件中的档案:

档案	描述	大小	格式	浏览次数
index.html		0Kb	HTML	262	检视/开启

在NCUIR中所有的数据项都受到原著作权保护.

社群 sharing

数据加载中.....