RFID based People-Object Direction of Pass Detection Ral Parada a , - PowerPoint PPT Presentation

RFID based People-Object Direction of Pass Detection Raúl Parada a , Joan Melià-Seguí b , c and Rafael Pous a Universitat Pompeu Fabra (UPF) a Departament de Tecnologies de la Informació i les Comunicacions (DTIC) Universitat Oberta de Catalunya (UOC) b Estudis d’Informàtica, Multimèdia i Telecomunicació (EIMT) c Internet Interdisciplinary Institut (IN3) 12th International Conference on Intelligent Environments London, United Kingdom - September 12-16, 2016

Customer behavior

Table of Contents 1 Introduction 2 People-Object Movement Detection Principle 3 Methodology 4 Experiments and Results 5 Conclusion & Future Work

Table of Contents 1 Introduction 2 People-Object Movement Detection Principle 3 Methodology 4 Experiments and Results 5 Conclusion & Future Work

Motivation & Research Objectives • Retailers do not know the hot-spots within their stores, to increase sales. • A study of different methods to correctly classify the people-object direction of pass. • A comparison of different state-off-the-art RFID antennas to increase the people-object direction of pass detection rate. • An extensive empirical analysis of people-object direction of pass with multiple tags and events. Raúl Parada 1/21 Introduction The proposed system Methodology Evaluation Conclusion

Motivation & Research Objectives • Retailers do not know the hot-spots within their stores, to increase sales. • A study of different methods to correctly classify the people-object direction of pass. • A comparison of different state-off-the-art RFID antennas to increase the people-object direction of pass detection rate. • An extensive empirical analysis of people-object direction of pass with multiple tags and events. Raúl Parada 1/21 Introduction The proposed system Methodology Evaluation Conclusion

Motivation & Research Objectives • Retailers do not know the hot-spots within their stores, to increase sales. • A study of different methods to correctly classify the people-object direction of pass. • A comparison of different state-off-the-art RFID antennas to increase the people-object direction of pass detection rate. • An extensive empirical analysis of people-object direction of pass with multiple tags and events. Raúl Parada 1/21 Introduction The proposed system Methodology Evaluation Conclusion

Motivation & Research Objectives • Retailers do not know the hot-spots within their stores, to increase sales. • A study of different methods to correctly classify the people-object direction of pass. • A comparison of different state-off-the-art RFID antennas to increase the people-object direction of pass detection rate. • An extensive empirical analysis of people-object direction of pass with multiple tags and events. Raúl Parada 1/21 Introduction The proposed system Methodology Evaluation Conclusion

Related Work (I) Using computer vision and sensors Ahmed et al., Design and Chung et al., Locomotive wireless implementation of a mobile based gate video recorder and recording system , 2014. passing detection method , 2015. Raúl Parada 2/21 Introduction The proposed system Methodology Evaluation Conclusion

Related Work (II) Only using RFID data Lee et al., A gate sensor for Nikitin et al., Phase based spatial construction logistics , 2008. identification of UHF RFID tags , 2010. Raúl Parada 3/21 Introduction The proposed system Methodology Evaluation Conclusion

Related Work (III) Also using RFID data Oikawa et al., Tag movement direction estimation methods in an RFID gate system. , 2009. Raúl Parada 4/21 Introduction The proposed system Methodology Evaluation Conclusion

Table of Contents 1 Introduction 2 People-Object Movement Detection Principle 3 Methodology 4 Experiments and Results 5 Conclusion & Future Work

Direction of Pass Principle • Based on the order in detecting labeled objects from the antennas • The object A is leaving when first read by Ant1 and last by Ant 2 • Or entering with opposite read order "Object B" Raúl Parada 5/21 Introduction The proposed system Methodology Evaluation Conclusion

Direction of Pass Principle • Based on the order in detecting labeled objects from the antennas • The object A is leaving when first read by Ant1 and last by Ant 2 • Or entering with opposite read order "Object B" Raúl Parada 5/21 Introduction The proposed system Methodology Evaluation Conclusion

Direction of Pass Principle • Based on the order in detecting labeled objects from the antennas • The object A is leaving when first read by Ant1 and last by Ant 2 • Or entering with opposite read order "Object B" Raúl Parada 5/21 Introduction The proposed system Methodology Evaluation Conclusion

RFID • Semantic indicators • Identification code (96-bit typically) • Antenna Port • Timestamp • Additional indicators • Received signal strength indicator (RSSI) • Radio frequency phase (PHASE) • Read Count Raúl Parada 6/21 Introduction The proposed system Methodology Evaluation Conclusion

RFID indicators • Extract RFID indicators from a passive RFID tag • The RSSI indicator shows a dynamic behavior of a passive tag Raúl Parada 7/21 Introduction The proposed system Methodology Evaluation Conclusion

RFID indicators • Extract RFID indicators from a passive RFID tag • The RSSI indicator shows a dynamic behavior of a passive tag Raúl Parada 7/21 Introduction The proposed system Methodology Evaluation Conclusion

Features’ generation • Generated RFID features from low and high-level indicators 1 BEAM: Positive and negative numeric identifiers to tag each antenna lobe depending on their position. 2 TIME: Interval of time in seconds between the first sample and the rest. 3 AV_TIME: Mean of the feature TIME in seconds from each antenna port indicator. 4 MINUS TIME: Interval of time in seconds between the last sample and the rest. 5 M1: Product of multiplying TIME per BEAM. 6 M2: Product of multiplying MINUS TIME per BEAM. Raúl Parada 8/21 Introduction The proposed system Methodology Evaluation Conclusion

Methods and Weights • Designed methods to detect direction of pass using uniquely RFID data 1 Slope: Based on a lineal model from the BEAM and TIME features. 2 Average Time: Based on the AV_TIME feature. 3 Momentum: Based on the M1 and M2 features. • Implemented weights to enhance the methods: 1 W1: It is the read count indicator. 2 W2: Difference of a given RSSI indicator value and the minimum among all the samples. 3 W3: Result of multiplying the weights W1 and W2. Raúl Parada 9/21 Introduction The proposed system Methodology Evaluation Conclusion

Table of Contents 1 Introduction 2 People-Object Movement Detection Principle 3 Methodology 4 Experiments and Results 5 Conclusion & Future Work

Experimental Setup • Two RFID antennas (with different directivity and gain) • Eight labeled-objects (two diverse passive tags) • RFID reader: • ThingMagic M6e (ETSI) • Read rate: 400 tags/second Raúl Parada 10/21 Introduction The proposed system Methodology Evaluation Conclusion

Experimental Setup PHASED ARRAY SYSTEM TILTED SYSTEM 4 1 2 1 3 2 • Two scenarios: 1 Phased array: Four beams generated Using a phase shifter 2 Tilted: Tilt antennas to project beams in opposite directions Raúl Parada 11/21 Introduction The proposed system Methodology Evaluation Conclusion

Tests • Passive RFID Tags • RFID Antennas • Web G2iL • 1x3 array • Frog 3D • 2x2 array • 1x1 air array • Different products • Reading mode • Material • Size • Sequential • Shape • Autonomous • Multitag • Antenna’s position • Multiple • No-tilted labeled-objects • Tilted • Phased • Multievent • Entering • Read Power • Leaving • 30 dBm • Static • 27 dBm Raúl Parada 12/21 Introduction The proposed system Methodology Evaluation Conclusion

Table of Contents 1 Introduction 2 People-Object Movement Detection Principle 3 Methodology 4 Experiments and Results 5 Conclusion & Future Work

RFID Antennas (I) Phased - RFID Antenna: 1x3 array - Read Power: 30 dBm Methods 75 Slope...No.weight Slope...W1 Slope...W2 Slope...W3 Accuracy [%] Average.Time...No.Weight 50 Average.Time...W1 Average.Time...W2 Average.Time...W3 Momentum...No.weight Momentum...W1 25 Momentum...W2 Momentum...W3 0 1 – 2 1 – 2 – 3 – 4 1 – 3 1 – 4 2 – 3 2 – 4 3 – 4 Combination of beams Raúl Parada 13/21 Introduction The proposed system Methodology Evaluation Conclusion

RFID Antennas (II) Phased - RFID Antenna: 2x2 array - Read Power: 30 dBm 100 Methods Slope...No.weight 75 Slope...W1 Slope...W2 Slope...W3 Accuracy [%] Average.Time...No.Weight Average.Time...W1 50 Average.Time...W2 Average.Time...W3 Momentum...No.weight Momentum...W1 Momentum...W2 25 Momentum...W3 0 1 − 2 1 − 2 − 3 − 4 1 − 3 1 − 4 2 − 3 2 − 4 3 − 4 Combination of beams Raúl Parada 14/21 Introduction The proposed system Methodology Evaluation Conclusion