以影像為基礎的食材辨識與智慧型冰箱管理系統

Image-based Food Recognition and Management System of Intelligent Refrigerator

陳君睿(Jun-Rui Chen)

指導教授 ： 劉震昌

國立暨南國際大學/科技學院/資訊工程學系/碩士(2011年)

摘要

在現今社會，由於科技的進步與環保觀念的推廣，消費者對於家電的要求已不僅是價格，更多為更好的使用者體驗與家電的附加功能，故智慧型家電的研發逐漸受到重視。本論文提出了一套以影像為基礎的食材辨識與智慧型冰箱管理系統，其包含了食材辨識與管理伺服器端與手持裝置的客戶端兩大系統。食材辨識與管理伺服器端則又可分成食材辨識、倉儲管理與食譜搜尋三個系統模組。我們擷取食材影像的色彩、紋理與長寬比特徵向量，搭配支持向量機，完成一套針對新鮮食材的影像辨識系統，達到了平均 96.9031% 的辨識率。倉儲管理系統則透過食材辨識對冰箱食材進行記錄，供使用者查詢份量與保存期限訊息。食譜搜尋系統提供了約一萬道中文食譜供使用者搜尋。手持裝置端則整合食譜管理與健康管理兩個系統模組，讓使用者可以隨時查詢食譜、規劃食譜、上傳個人食譜與管理記錄個人健康資訊。

關鍵字

智慧型冰箱 ； 食材影像辨識 ； 支持向量機

並列摘要

In recent years, with the development of scientific and technological progress, the research and development of home appliances is getting more attention because customers require better user experiences and additional features of home appliances than prices. In this thesis, we propose an Image-based Food Recognition and Management System of Intelligent Refrigerator. This system includes two subsystems: a food recognition and management server and a handheld device client. The food recognition and management sever has three system components which include food recognition、warehouse management and recipes search. We extract color feature、texture feature and aspect ratio feature of food images, then apply support vector machine to build a fresh food image recognition system. The average recognition rate is 96.9031%. By using food image recognition, the warehouse management system can automatically record the name, weight and expire date of fresh food which user can inquire these data with this system. The user can search recipes among 10,000 Chinese recipes. The handheld device client has two functionalities including recipes management and health management, so user can search recipes、get recipes plan、upload personal recipes and record personal health information.

並列關鍵字

intelligent refrigerator ； food recognition ； support vector machine

參考文獻

[1] A. M. McIvor, “Background Subtraction Techniques”, Proc. of Image and Vision Computing, 2000.

[2] A. Leykin, M. Tuceryan, “Automatic Determination of Text Readability over Textured Backgrounds for Augmented Reality Systems”, Third IEEE and ACM International Symposium on Mixed and Augmented Reality, pp.224-230, 2004.

[3] D. G. Lowe, “Distinctive Image Features from Scale-invariant Keypoints”, International Journal on Computer Vision, 2004.

[4] G. Shroff, A. Smailagic and D. P. Siewiorek, “Wearable Context-Aware Food Recognition for Calorie Monitoring”, IEEE International Symposium on Wearable Computers, pp. 119-120, 2009.

[5] K. Aizawa, G.C. de Silva, M. Ogawa, Y. Sato, “Food Log by Snapping and Processing Images”, 16th International Conference on Virtual Systems and Multimedia (VSMM), pp.71-74, 2010.