視覺式計人次暨性別辨識系統-以數位廣告看板為例

A Real Time Visual People Counting and Gender Classification System-for Digital Signage Application

林柏宏(Bo-Hon Lin)

指導教授 ： 謝禎冏

大同大學/電機資訊學院/資訊工程學系所/碩士(2009年)

摘要

在本篇論文中，我們設計出一個新的即時計人次且具記憶系統，並且更進一步的辨識潛在客戶性別，主要應用在展覽會場與購物中心，其功能是讓廣告主知道目前放映的廣告，吸引了多少人前來觀看，並且觀看的族群中，男女生比例的多寡。首先，我們利用Adaboost偵測人臉，再利用一些條件濾除在廣告看板前非人或非真正在觀賞廣告的人。為了算出精確的觀看廣告看板的人流數目，我們擷取出一些特徵，像是人臉紋理、軀幹紋理、質量中心等，最後將這些特徵存入資料庫做相似度比對。在資料庫更新上，我們採用Least Recent Used的機制，刪除不常使用的特徵資料。在性別辨識的功能上，我們擷取的辨識特徵主要為頭髮及人臉大小，並且計算比例，依此判斷目標是否為男或女。在實驗結果中，人數計次和男女辨識的準確率分別為95.55%和94.1%，而執行效能在一般個人電腦上可達15~20fps。

關鍵字

電腦視覺 ； 人臉偵測 ； 性別辨識 ； 人數計次 ； 數位看板

並列摘要

In this paper, we designed a real time person counting/memorizing system which could further distinguish the gender of potential customers. Firstly, we employ Adaboost to detect possible human faces and utilize some criteria to filter out the wrong ones. For each detected person, face, torso, and mass center are recorded in database for people matching/tracking. The least recently used record would be deleted if database is full. Regarding the gender-classification function, the ratios of hair to facial size and beard to chin area are calculated to determine the sexes. From the experimental results, person counting and gender classification each has an accuracy of 95.55% and 94.1%, respectively. The execution efficiency on personal desktops can reach as high as 15-20 fps.

並列關鍵字

computer vision ； face detection ； gender classification ； people counting ； Digital signage

參考文獻

[1] S.-Y. Cho, T. W. S. Chow, and C.-T. Leung, "A Neural-Based Crowd Estimation by Hybrid Global Learning Algorithm," IEEE Transactions on Systems, Man, and Cybernetics - Part B, Vol. 29, No. 4, August 1999.

[3] S.-Y. Cho, T. W. S. Chow, and C.-T. Leung, "A Neural-Based Crowd Estimation by Hybrid Global Learning Algorithm," IEEE Transactions on Systems, Man, and Cybernetics - Part B, Vol. 29, No. 4, August 1999.

[5] N. Paragios, and V. Ramesh, "A MRF-based Approach for Real-Time Subway monitoring," Computer Vision and Pattern Recognition, 2001.

[6] D. Roqueiro and V. A. Petrushin, “Counting People using Video Cameras,” MDM/KDD'06, August 20, 2006.

[7] A. B. Chan, Z. S. Liang, and N. Vasconcelos, “Privacy Preserving Crowd Monitoring: Counting People without People Models or Tracking,” Proc. IEEE Conf. on Computer Vision and Pattern Recognition, June 2008.