Participatory Sensing in Commerce: Using Mobile Camera Phones ...

Participatory Sensing in Commerce: Using Mobile Camera Phones ...

PARTICIPATORY SENSING IN COMMERCE: USING MOBILE CAMERA PHONES TO TRACK MARKET PRICE DISPERSION Nirupama Bulusu Portland State University Chun Tung Chou, Sali Kanhere, Yifei Dong, Shitiz Sehgal, David Sullivan, and Lupco Blazeski University of South Wales From UrbanSense 2008 Presented by: Travis Cossairt April 4th, 2011 EEL 6788

OVERVIEW Introduction Related Work Challenges Proof of Concept Systems: PetroWatch: Automated Fuel Price Collection Mobishop: Semi-automated Scanning of Receipts Conclusion Discussions

INTRODUCTION Price Dispersion refers to price difference of a homogenous good across different vendors, e.g. gas price, grocery items, etc. Attributed to several causes including the consumer search cost incurred in collection price information (i.e. takes time to look up and compare prices) Good reasons to collect this info, notwithstanding empowering consumers to negotiate better prices

INTRODUCTION (CONT.) Existing consumer communities already tracking price dispersion (manually): Hong Kong housewives divided themselves into teams and manually copied prices of selected staple grocery items and uploaded to a website, led to newspaper to advertise on its website Fuel prices are collected manually in several countries by volunteers and uploaded to websites Manual process however is cumbersome,

error-prone, and not up-to-date Authors propose their vision of using participatory camera phone sensing to solve some of these problems RELATED WORK Participatory Sensing systems used in several domains including: Collecting health info (DietSense [1]) Intelligent Transportation (Traffic Sense [2])

Air-Quality Monitoring (Sensing Atmosphere [3]) Authors have prepared two proof of concept systems in the domain of commerce They suggest that were first to apply to commerce Reuters Market Light system provides farmers with up-to-date crop prices [4], but is distinct from this work in that the authors focus on empowering consumers and collecting information from them CHALLENGES Data from camera phones not in a consistent

format, thus hard to aggregate pricing info across different retailers (vs. well formed web XML) Larger # of goods/consumers make it difficult to collect information into a single database (thus most apps/system focus on a single good) Computer vision challenges extracting price info Having optimal positioning of camera phones for image capture MORE CHALLENGES

Inherent challenges in all participatory systems: Incentivizing participation Need to lower technical and any monetary barrier System should be simple and automated Take advantage of free channels like WiFi Offer information reward to contributers Security, privacy, and data reliability

Need to protect not only location for privacy, but also shopping behavior Integrity of data needs to be ensured such that no bogus data is contributed (e.g. retailer poisoning data to show competition is higher) Authors expect to build upon other solutions being developed in the long term, but not in this paper PROOF OF CONCEPT SYSTEMS

Two proof of concept systems built to show that processing and deliver product pricing info via mobile camera phones can be accomplished: PetroWatch Automates collection of fuel prices by taking photographs of roadside fuel price boards, processed by computer vision, and mapped to GIS data of stations in the backend MobiShop

Photograph store receipts, processed by OCR, and stored into central database for store/product comparisons PROOF OF CONCEPT SYSTEMS ARCHITECTURE Both systems built using similar client-server architectures Operate in two modes: Price collection, User Query PROOF OF CONCEPT DETAILED PAPERS

The two applications described here are further detailed in their own papers: Automatic Collection of Fuel Prices from a Network of Mobile Cameras [5] MobiShop: Using Mobile Phones for Sharing Consumer Pricing Information [6] PETROWATCH: AUTOMATED FUEL PRICE COLLECTION Automated collection of fuel prices by triggering mobile phones to photograph fuel

price boards when users approach service stations PETROWATCH: AUTOMATED FUEL PRICE COLLECTION Central server implements the computer vision algorithms for processing the images and extracting the fuel prices Relies on GIS database and GPS location to derive what station the user is at Also relies on unique fact that each brand has a specific color for its fuel price board PETROWATCH: AUTOMATED FUEL

PRICE COLLECTION PETROWATCH: AUTOMATED FUEL PRICE COLLECTION Post-processing techniques used in some cases to narrow search when other similarly colored objects in photograph (e.g. the sky vs. Mobil fuel price board) Color thresholding, color histogram, etc. used to get precise board location

Image cropped to contain only the board, and normalized to standard size/resolution Feedforward Backpropagation Neural Network algorithm used to classify price numeral characters PETROWATCH: AUTOMATED FUEL PRICE COLLECTION PETROWATCH: AUTOMATED FUEL PRICE COLLECTION PETROWATCH: AUTOMATED FUEL PRICE COLLECTION

Prices stored in a database, and linked to GIS road network database populated with the service station locations Server updates fuel prices for appropriate station if current image has a new timestamp History recorded to allow analyzing pricing trends MOBISHOP: SEMI-AUTOMATED SCANNING OF RECEIPTS Users photograph actual store receipts, and OCR technology to extract pricing and

product info MOBISHOP: SEMI-AUTOMATED SCANNING OF RECEIPTS Uploaded along with location/time info, and interfaces with a GIS street map for viewers Using native Symbian OS 9.2 OCR engine, achieved about 60% accuracy of items Allows users to manually edit text to fix the rest Also ability to delete personal info like credit card details MOBISHOP: SEMI-AUTOMATED

SCANNING OF RECEIPTS Implemented in JavaME for portability on a Nokia N95 8GB phone Interfaces with external GIS library (J2MEMap) to highlight store locations on street map for navigation Server side also written in Java/Tomcat

CONCLUSIONS Authors explored participatory camera sensing for tracking price dispersion in offline markets (fuel stations, retail product prices) Claimed to be first to apply participatory sensing in commerce domain System could be extended to help consumers track spending habits, receive promotions, and other track other homogenous goods DISCUSSION

Strengths Two simple and interesting applications of participatory sensing in the domain of commerce Weaknesses PetroWatch very dependent on assumption that each brand has unique colored fuel price board these days many are digital with black background!

Also must assume stations have just Reg,Mid,Premium MobiSense depends on items being entered that are purchased, thus no prices tracked for items not purchased or popular (or maybe too expensive!) 60% OCR accuracy still requires a lot of manual entries/editing (authors admit future work here) DISCUSSION CONT. Applications seem to be a good start, but never fully explored or made into complete system thats useful Fail to describe the usage of the data, e.g. all the fuel prices uploaded to server, what next?

Seems to focus on data collection only in this paper scope ADDITIONAL REFERENCES [1] S. Reddy, A. Parker, J. Hyman, J. Burke, D. Estrin, and M. Hansen, Image Browsing, Processing, and Clustering for Participatory Sensing: Lessons From a DietSense Prototype, in Embedded Networked Sensors, Ireland, June 2007. [2] P. Mohan, V. Padmanabhan, and R. Ramjee, TrafficSense: Rich

Monitoring of Road and Traffic Conditions using Mobile Smartphones, Microsoft Research Technical Report, MSR-TR-200859, Apr 2008. [3] E. Paulos, R. Honicky, E. Goodman, Sensing atmosphere, In: Workshop on Sensing on Everyday Mobile Phones in Support of Participatory Research, Sydney, Australia, Nov 2007. [4] D. Grammaticas, Text messages empower poor farmers, BBC News, 6 May 2008.

[5] Y.F. Dong, S. Kanhere, C.T. Chou, and N. Bulusu, Automatic Collection of Fuel Prices from a Network of Mobile Cameras [6] Shitiz Sehgal, Salil S. Kanhere, Chun Tung Chou, MobiShop: Using Mobile Phones for Sharing Consumer Pricing Information

Recently Viewed Presentations

  • Chapter 13

    Chapter 13

    Writing Lewis Structures HCCH Writing Lewis Structures HCCH Writing Lewis Structures BeCl2 Steps for Writing Lewis Structures If you cannot write a stable structure for the molecule using rules 1 - 3 add or remove unshared pairs to/from the central...
  • Journal - Yeg Lit

    Journal - Yeg Lit

    Concrete details. Discrete information used to describe something. Helps the reader visualize or comprehend an idea. Typically factual details - you can't argue with them . Rely mainly on adjectives . Can be helpful, but not always the strongest for...
  • ISDN Box - Chipps

    ISDN Box - Chipps

    ISDN Equipment. At the end user site an ISDN line only requires a single box, but this box performs several functions. The type of box also depends on whether the box is ISDN aware or not. Specialized ISDN terminals are...
  • Cost Management and Strategy: An Overview

    Cost Management and Strategy: An Overview

    Ethical Issues in Cost Allocation. An ethical issue arises when costs are allocated to products or services that are produced for both a competitive market and a public or governmental entity. The latter often purchases on a cost-plus basis creating...
  • Edru Research Day

    Edru Research Day

    Working within such tradition, we can argue that Samantha has made choices which are not communicatively effective. In other words, the meanings she could have construed but did not are just as significant as what she did chose. Genre -...
  • Managed Extensibility Framework (MEF) @ Bonn-to-Code

    Managed Extensibility Framework (MEF) @ Bonn-to-Code

    Offizielle Beschreibung "The Managed Extensibility Framework (MEF) is a new library in the .NET Framework that enables greater reuse of applications and components.
  • Programmering - fra instruktion til modellering

    Programmering - fra instruktion til modellering

    Organization of conferences. Each conference has a program committee . with 20-30 PC-members lead. by . 1-2 . two PC-chairs. All PC-members are well-known researchers within the area covered by the conference (typically appointed by a steering committee in cooperation...
  • u - Amazon S3

    u - Amazon S3 19/09/2017. Personal Profile. Search. Apply. Research. Select 'Profile' from the top bar and work your way through each tab by entering the relevant details. Once all sections have been completed and turned green you will then be able to...