基於機器學習的Beacon室內定位應用於智慧零售系統;ML-based Beacon Indoor positioning for Smart Retail System

NCUIR > College of Electrical Engineering & Computer Science > Executive Master of Computer Science and Information Engineering > Electronic Thesis & Dissertation > Item 987654321/89752

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/89752

Title:	基於機器學習的Beacon室內定位應用於智慧零售系統;ML-based Beacon Indoor positioning for Smart Retail System
Authors:	林琮皓;Lin, Tsung-Hao
Contributors:	資訊工程學系在職專班
Keywords:	智慧零售;遞迴式機率神經網路;藍芽Beacon;室內定位;Smart Retail;Recursive Probabilistic Neural Network;Bluetooth Beacon;Indoor Positioning
Date:	2022-07-25
Issue Date:	2022-10-04 11:58:28 (UTC+8)
Publisher:	國立中央大學
Abstract:	零售業導入智慧零售的需求日益增長，但導入無人商店的智慧零售缺乏了互動性，也因此阻礙了智慧零售的發展。為了讓顧客不僅是體驗智慧零售而來，提升購物體驗則成了最大的課題，這也因此影響了零售業導入智慧零售的意願。本論文中提出一套基於機器學習的Beacon室內定位智慧零售互動平台，利用藍芽Beacon訊號進行特徵擷取，再透過遞迴式機率神經網路進行位置的辨識，搭配樹莓派作為電子紙互動平台閘道器，並將辨識結果傳至樹莓派控制電子紙並與顧客進行互動。透過MQTT作為樹莓派與PC之間的通訊方式，並使用PSO結合RPNN進行Beacon訊號的辨識，其平均辨識準確度可達到97.56%，與LSTM相比其效果提升6%。本論文規劃一套完整的智慧零售系統，使用藍芽Beacon結合機器學習提升定位準確度，並利用電子紙平台與顧客進行互動，以藍芽及樹莓派的應用降低智慧零售導入的成本，實現零售業的數位轉型。;There is a growing demand for smart retail in the retail industry, but the lack of interactivity in unmanned stores has hindered the development of smart retail. In order to let customers not only experience smart retailing, improving the shopping experience is the biggest issue, which affects the willingness of the retail industry to implement smart retailing. In this paper, we propose a machine-learning-based Beacon indoor location-based smart retailing interactive platform, which uses Bluetooth Beacon signals for feature acquisition, and then uses a recursive probabilistic neural network for location identification, and Raspberry Pi as the gateway to the e-paper interactive platform then transmits the identification results to Raspberry Pi to control the e-paper and interact with customers. By using MQTT as the communication method between Raspberry Pi and PC, then using PSO combined with RPNN for Beacon signal recognition, the average recognition accuracy can reach 97.56%, which is 6% better than LSTM. In this paper, we plan a complete smart retail system, using Bluetooth Beacon combined with machine learning to improve the positioning accuracy, and using the e-paper platform to interact with customers, using Bluetooth and Raspberry Pi to reduce the cost of smart retail implementation and realize the digital transformation of the retail industry.
Appears in Collections:	[Executive Master of Computer Science and Information Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	57	View/Open

社群 sharing

Loading...