Keywords

1 Introduction

Preventing illegal resale, such as ticket scalping [1–3], is socially required. Because illegal resale is a critical problem for both ticket users and event companies, strict steps are followed at popular events in Japan from ticket application up to entry admission, as shown in Fig. 1. In advance of an event, ticket users take the following steps: ticket application through fan club membership (step 1) and ticket purchase according to notification of winning (step 2). Tickets are not sent to users to restrain resale. On the event day, attendants take the following steps at the event venue: ticket winning identification by fan club membership cards (step 3), identity confirmation by visual inspection with ID cards or documents (step 4), and admission procedure including ticket issuing (step 5). Step 4 is a time-consuming task for venue attendants because fake IDs are available together with tickets through online auction websites. It is also stressful because ticket users feel uncomfortable when kept waiting.

Fig. 1.
figure 1

Example of current ticketholder identification procedure

Full size image

The key points in identification are to verify identities efficiently and to prevent people from impersonating others.

2 Methods

2.1 Identification System Using Face Recognition Software

We have developed an identification system using face recognition software as a way to improve confirmation efficiency and to prevent people from impersonating others. As shown in Fig. 2, the system registers facial photos when tickets are applied for in Step 1, and it verifies identities by using face recognition software rather than confirming with ID cards in Step 4.

Fig. 2.
figure 2

System’s ticket purchaser identification procedure

Full size image

Our system performs the following steps to identify ticket applicants and ticket holders:

  • Step 1: People applying for tickets register their membership information as well as their facial photo. At that time, they will read and agree to the privacy policy in effect regarding the handling of the photo and other personal information and the verification of their identity on the day of the event. In the same way as for an ordinary ID photo, the registered facial photo is a clearly visible frontal image taken against a plain background. The face must not be obstructed by a hat, sunglasses, mask, scarf, or the like, or by things like excessively long hair or a flashed peace sign.

  • Step 2: Successful applicants are notified in the usual manner.

  • Step 3: Successful applicant identities are confirmed by using a membership card reader in the usual manner.

  • Step 4: At the event, the attendant uses face recognition software to confirm that the photo taken at the time of application and the registered photo show the same person.

  • Step 5: The admission procedure is carried out in accordance with the face authentication results.

2.2 Face Recognition Software

The face recognition software the system uses is the internationally reputable commercial product NeoFace [4, 5]. It was implemented in a commercially available AGT10 tablet terminal with a rear view camera [6]. All applicants’ facial image information is copied to the tablet terminal in advance, and the terminal alone performs the face recognition process.

2.3 Confirmation Procedure

An event attendant performs the following confirmation procedure using the equipment shown in Fig. 3, comprising a card reader, display monitor, and tablet terminal implemented with face recognition software:

Fig. 3.
figure 3

Event venue equipment

Full size image
  1. (1)

    Attendees’ membership cards are placed on the card reader, and the monitor screen confirms the attendees are successful applicants. The screen displays the face images that were registered at the time of application.

  2. (2)

    The attendant explains the identification through a face recognition process to the attendees and instructs them where to stand in front of the terminal.

  3. (3)

    The attendant executes the face recognition process using the terminal to confirm the attendees are those who applied for the tickets.

  4. (4)

    If identification is confirmed, the attendee is admitted entry.

  5. (5)

    If identification is not confirmed, the face recognition process is repeated, or identity is confirmed by direct visual inspection.

3 Results and Discussion

3.1 Identification at a Concert Venue

The system was utilized for a July 26, 2014 pop music concert at Nissan Stadium in Yokohama. The equipment was installed in 120 locations just behind the baggage inspection site to perform face recognition for 50,324 attendees over two days, as show in Fig. 4. The weather was mostly sunny, but the area became dark temporarily due to a thunderstorm. Face recognition was performed only for ticket applicants and not for people attending with them. The average accuracy of face recognition was 90 %. The recognition failed when the people had their eyes closed, were not looking directly forward, or had hair covering their face. There were also cases where darkness due to the thunderstorm was a factor. The confirming step took 6 s on average or 7 s if we included cases where recognition was not achieved. This was 30 % more efficient than visually confirming identification through comparison with conventional ID cards, the time for which rose to 10 s. No cases of people impersonating others were reported for this event.

Fig. 4.
figure 4

Attendees being identified

Full size image

3.2 Preventing People from Impersonating Others

People purchasing tickets at websites were well aware that the registered face images of ticket applicants would be matched with the facial images of people attending the event when they entered the venue. Under these conditions, there were no reports about people attempting to impersonate others at the event. The system’s performance has been widely reported in the mass media [7, 8]. In addition to the aforementioned pop music concert, it has been used for 26,859 people at the Saitama Super Arena on December 24–25, 2014, for 33,434 people at Fukuoka YAFUOKU! Dome on April 4–5, 2015, and for 38,563 people at Shizuoka Stadium ECOPA on July 31–August 1, 2015. In fact, since the aforementioned pop music concert, it has been used more than 20 times for large scale events [9]. No cases of people attempting to impersonate others were reported for any of these events. This is indicative of the system’s effectiveness in improving equity in ticket purchasing and deterring or preventing illegal resale.

3.3 Making Verification More Efficient

The system achieved 30 % more efficiency than a visual identification with a conventional ID card. It also reduced the psychological workload for the event attendants. Most of the attendants were part-time workers who had to identify 500 to 1,000 people per day visually. Verbal exchanges with attendees and other factors put a high psychological workload on the attendants, and many of them said they likely would not do such work at future events because of these exchanges and related factors. According to the event organizers, the identification by a face recognition system makes it easier for them to find part-time attendants who will continue to do such work at future events.

3.4 Future Issues

To make the system’s identity confirmation process more efficient, we should consider ways to improve its operating environment and face recognition method. Its operating environment can be improved by installing lighting to compensate for insufficient lighting at the site. We could also make the system more efficient by finding ways that would improve the understanding and cooperation of users. There have been cases at event sites where attendees’ photos were taken, but their identity could not be confirmed because they had their eyes closed, because they were not directly facing the camera, or because their hair was obstructing their face. The problem was often compounded because the attendant was unable to give the attendees a good explanation as to why their identity could not be confirmed. Providing prior information relevant to face recognition, at the ticket application time or other times, would enable facial photos to be taken appropriately. In the future, attendee understanding can be expected to increase as the face recognition process and systems such as ours become more widespread. However, event attendants will need to explain to attendees more effectively how their photos taken on the day of the event will be handled to alleviate their concerns.

We plan to study the possibilities of introducing a “walk-through” system as a way to improve face recognition [10]. This would be a system where people are photographed as they approach the system equipment head on and be admitted entry if facial recognition succeeds. Having people photographed as they approach would save them from having to stop to have their photos taken and thus reduce waiting time. We will attempt to develop a practical way in which this can be done.

4 Conclusion

We have developed an identity confirmation system using face recognition software.

The system was proven effective for preventing illegal resale by confirming 50,324 attendees at a large concert of a popular music group. The average accuracy of face recognition was 90 %. The average time for identity confirmation was 7 s per person including guidance to ticket holders, which succeeded in decreasing identity confirmation time by 30 % using visual inspection as well as in reducing the psychological workload of venue attendants. To further streamline the identification procedure, in the future, we plan to improve the performance by introducing ways to explain the procedure to users more clearly and also by introducing a “walk-through” system.