How to Reduce Passengers Waiting Times in Video People Counter

- Aug 31, 2017-

Long waiting times make airports passengers feel very sad and upset. And long queues for checkpoint make airports lose money. Urgent passengers at the security checkpoint may have no interest at retail sales. People that wait more, spend less.

Conversely, cutting average security waiting time by one minute may entail higher revenues of USD 1.5 to 2.3 million, as Passenger Flow Terminal reported on a case study of passenger flows conducted at Toronto Pearson International Airport in September 2016. This correlation seems also to apply to smaller and mid-size airports. Federico Cabrera, CCO at Aeropuerto Internacional de Carrasco in Uruguay’s capital Montevideo confirms: “Shortly after we had installed a video people counter, we could optimize staff planning and hence reduce waiting times at security checkpoint. This led to a remarkable increase in retail sales of up to 10’000 $ in just three hours”.

A Broad Range of Technological Options

In recent years, the array of technologies to address the issue has expanded along with the awareness of the problem. Many different names are used to describe technologies to measure waiting times, such as queue measurement systems, passenger flow analysis and people flow monitoring. Typically, all the available technologies count people in the relevant area (e.g. at security checkpoint or even in check-in/ticketing areas) and measure their dwell times, based on which an algorithm calculates the resulting waiting times. The differences between the various approaches lie in cost, installation and maintenance efforts, accuracy, reliability, and scalability:

  • Wi-Fi/Bluetooth: Depending on passengers having Wi-Fi/Bluetooth enabled on their smartphones or other devices, these features can be used to estimate waiting times. Usually, sample rates of such systems are rather low, since only a fraction of the enabled devices can be detected.

  • Infrared: At the start and end of a designated queuing area passenger are counted when crossing a threshold and thereby breaking a beam of light from a sensor. The waiting time is measured based on the number of passenger in between.

  • Thermal cameras: Thermal cameras measure waiting times based on heat emitted by passengers. Since thermal cameras detect heat in general, he measurements are prone to error by other sources of heat.

  • Conventional video based systems: A specific software processes the video output from conventional cameras and calculates the waiting time. Using already existing cameras, installation is simple but can interfere with other video based applications and increase the total lifetime costs considerably due to the required network and computing capacities. As with all technologies based on two-dimensional images (monovision), bright sunlight, shadows, and other external influences negatively affect the accuracy.

  • 2D Camera: Typically mounted on the ceiling, 2D sensors have a built-in camera (monovision). The sensor differs from regular CCTV cameras by directly processing the image on the sensor itself to count people and measure the dwell time. Thus, network load and required server capacity is significantly reduced. An additional software combines various sensors to cover large areas and to measure the waiting time. As with all monovision technologies, light changes, shadows, and other external influences negatively affect the accuracy.

  • 3D Binocular Cameras: Typically mounted on the ceiling, 3D passenger counting sensors have two inbuilt cameras (stereo cameras) to calculate a 3D depth image. A software on the sensor performs the image processing directly on the people coutner to count people and measure the dwell time. An additional software combines various sensors to cover large areas and to measure the waiting time. 3D video counting system is the best option against light changes, shadows etc. 

Accuracy Gap Between Different Optical Systems

Considering the major challenges for airports, optical systems seem to have an edge over other technologies. Only optical solutions enable users to constantly receive real-time data along with live views of the relevant areas. But not every optical technology meets the expectations to the same degree: conventional video camera based approaches arise from the desire to combine existing video cameras (CCTV) with video analytics software and hence avoid investments in new equipment. But they fall short of the required accuracy and reliability, as the image processing in monovision, i.e. one camera, is limited per se. For example, if a passenger stands closely in front of another passenger form the camera’s perspective, only one person instead of two are counted and the waiting time is distorted. Furthermore, the image processing occurs on external devices and requires both considerable network and computing capacity, which increases total lifetime costs significantly. 2D coutning sensors enable direct image processing, but feature the same accuracy/ reliability short-comings attributed to monovision per se and only satisfy the requirements in well lighted environments. All in all, systems based on 3D people counting sensors deliver the most complete package, when it comes to measuring waiting times at highly frequented and complex airport sites with demanding passengers.

A 3D System in Action

A glimpse at a practical example sheds light on the question why binouclar caemras based system achieve the highest level of accuracy, reliability, scalability while keeping the total lifetime cost low. When entering the covered area, every passenger is counted, tracked and allocated to a queue according to an “automated queue detection” algorithm. Passengers that are not queuing, are excluded from the measurements. Each of the deployed devices combines two image sensors to see three-dimensionally – very similar to the human eye; two “eyes” identify pixels that correspond to the same point in an observed scene to determine the 3D position of this point via triangulation.

3D sensors are independent from signal-emitting devices and highly robust against external influences, such as shadows, light changes and heat emissions. Numerous tests have substantiated the accuracy of the Xovis system measuring the desired KPIs.