挖掘潛在顧客之商業智慧系統－以影像辨識技術為基礎

A Business Intelligent System for Mining Potential Customers－Based on Image Recognition Technology

呂彥輝(Yan-Hui Lu)

指導教授 ： 蔡玉娟

國立屏東科技大學/管理學院/資訊管理系所/碩士(2013年)

https://doi.org/10.6346/NPUST.2013.00184

摘要

現今網路發達的社會中，企業不斷地推出新商品來增加熱門度，更多企業利用以往累積的客戶資料來加強競爭力，他們從這些龐大的資料中尋找潛在顧客，就是未來可能成為買家的人；潛在顧客一直以來是受到各方企業關注的一個議題，他們對於市場來說就像是幽靈人口，有消費能力卻無消費紀錄可以追蹤，以致其購物動機不易察覺。大多企業、商家皆利用會員制度去發掘潛在顧客的市場，但對於不願填寫會員資料的人口依然是無法有效的去發掘。

本研究以影像辨識處理技術為基礎，針對沒有填寫會員資料的人口建置一套商業智慧系統－「挖掘潛在顧客之商業智慧系統－以影像辨識技術為基礎」，包括：(1)移動物體找尋－異物入侵偵測，找尋並擷取出畫面中的移動物體，以供後續辨識用；(2)人臉判斷模組－將畫面經由YCbCr膚色分類處理減少雜訊，再用Haar-like矩形特徵方法擷取出人臉範圍；(3)顧客特徵辨識、比對模組－確定移動物體是人之後，交由Scale Invariant Feature Transform尺度不變特徵轉換(SIFT)進行臉部辨識取得特徵點，找出該名顧客是否有符合資料庫中的樣本；(4)辨識結果流程－系統會依照比對結果判別是否有相似樣本，有的話服務人員便會更新資料庫，若無則會為資料庫新增該名顧客資料，藉此達到潛在顧客的紀錄與管理。

藉由系統對顧客臉部影像與購買記錄的管理功能，服務人員便可以依照系統所提供的購買紀錄針對不同種類的顧客提供服務，提升顧客滿意度及忠誠度，藉以用影像技術達成挖掘潛在顧客。本研究也藉由實驗人臉影像辨識技術應用於商業智慧系統之可行性，供後續研究做為參考。

關鍵字

臉部辨識 ； 商業智慧 ； 潛在顧客

並列摘要

Nowadays, with the development of Internet in our society, enterprises not only continuously issuance new products to increase their popularity but also use their own members’ profiles to enhance their competitiveness. They are looking for potential customers whom might be their future buyers from the huge mass of data. And these potential customers have always been an issue that all enterprises pay highly attention to. To every enterprise, the potential customers are like phantom population for the market because they already have consumption capacity but their consumption records are few to be analyzed so it is hard for us to sense their shopping motivations. Most enterprises and companies use membership system to find potential customers; but to those who are unwilling to give their personal information when merchandising, it is still a problem.

In this study, we use image recognition processing technology as the foundation to build a business intelligence system named “A Business Intelligent System for Mining Potential Customers – Base on Image Recognition Technology” for those people that do not fill out membership applications. The system include: (1) Finding moving objects – Foreign objects intrusion detection. This part is to find moving objects from images to provide identification module. (2) Face detection module – Use YCbCr skin color classifying technology to process images to reduce noises, then use Harr-like rectangle features to capture face area. (3) Customers features recognition and matching module – After confirming the moving object is a human being, this process will use Scale Invariant Feature Transform technology to check if the moving object has matched the samples from the database. (4) Procedure of identification results – The system will identify the similar samples according to the matching results. If it is matched, staff will update their customers’ data; if it is not matched, then staff can also add this new customer to database and let this customer become a new member. By this way, we can achieve in recording and managing potential customers.

With the system’s customer face image and purchase records management functions, staff can offer different services for different customers according to the purchase records in order to increase customers’ satisfaction and loyalty to the enterprises. Furthermore, recognition technology can help mining potential customers successfully. In this study, we experiment the feasibility of face image recognition technology applied to business intelligence system and this thesis is provided for follow-up research for reference.

並列關鍵字

Face Recognition ； Business Intelligence ； Potential Customers

參考文獻

[2] 吳明衛，「自動化臉部表情分析系統」，國立成功大學資訊工程系碩士論文，2003年。

[1] B. Shoushtarian, H. E. Bez, “A practical adaptive approach for dynamic background subtraction using an invariant colour model and object tracking,” Pattern Recognition Letters, Vol. 26, No. 1, 2005, pp. 5-26.

[2] C. P. Papageorgiou, M. Oren, T. Poggio, “A general framework for object detection, ”Sixth International Conference on Computer Vision, 1998, pp. 555-562.

[3] D. G. Lowe, “Object recognition from local scale-invariant features,” In International Conference on Computer Vision, Corfu, Greece, 1999, pp. 1150–1157.

[4] D. G. Lowe, “Distinctive image features from scale-invariant keypoints,” International Journal of Computer Vision, Vol. 60, No. 2, 2004, pp 91-110.