about 1 month ago
Automatic Passenger Counting - An Overview
Systems for Automatic Passenger Counting (APC) automatically record the number of people in public transport vehicles. This article is intended to give an overview of the different types of automatic passenger counting systems. The strengths and weaknesses of different technical approaches are examined.
Why Passenger Counting is Important
A demand-oriented design of public transport systems requires a thorough analysis of traffic flows. To ensure sufficient supply of public transport it is common practice to perform passenger surveys repetitively over time.
Typical questions to be solved in this context:
- Number of people inside the vehicle (utilization)
- Fluctuation at each stop
- Ticket type
- Entry and exit point
- Used types of transportation
Since such surveys must be carried out in a large area and over a long period of time in order to draw significant conclusions, such surveys are often accompanied by a great financial burden.
To make matters worse, even with high survey density, individual aspects that are not permanent or restricted to a small geographic area can not be captured.
To tackle these problems, there is an effort to automatically and continuously answer as many of the above questions as possible.
Over time, various approaches have been developed. Each of which with different strengths and weaknesses.
Different solutions to automatic passenger counting (APC)
Beam sensors inside the doors
A relatively simple way to measure the number of people in the vehicle is to install a light barrier in each door. Depending on the deviation from the undisturbed state, the system recognizes whether a person has entered or left the vehicle. Measuring in and outflow at each stop the total number of passengers can be measured during a journey the vehicle.
In such systems, the weakness is that simultaneous entry and exit of multiple people is often incorrectly recognized. Likewise, bicycles or strollers are a problem. Additionally, one has to account for the extra sensor cost.
It should not be forgotten that any misjudgment of entry and exit has a negative impact on the accuracy of the total number of people. Over the duration of a trip, the error can increase steadily.
Pressure sensors inside the seats
One can also measure occupancy using pressure sensors in the seats. If a seat is occupied, the sensor sends a signal to the associated control unit.
Disadvantage of this method is that only the load of the seats is considered. However, especially at peak times majority of passengers are standing. Furthermore, the cost of such a system is relatively high because each seat must be equipped with a sensor.
Measuring the weight of the train
Each person increases the total weight of the train. Measuring the total weight of the train and subtracting the weight when being empty, one obtains a rough estimate of the occupancy of the train.
Measuring mobile signals
Most people carry mobile devices such as smartphones which are emitting signals such as WiFi, Bluetooth or LTE. Measuring these signals one can infer the number of people inside the vehicle.
However, many people carry multiple devices or no devices at all, which reduces the accuracy.
Ticket applications on smartphones of customers can be used as source of information for passenger counting. However in rural areas as well as in tunnels internet connection and GPS location might be bad or distorted making it hard to measure the occupancy of single vehicles. As long as there is analog ticket distribution the information obtained from ticket apps remains incomplete.
At entrances and exits with payment barriers
Payment systems like the one in the image below, can capture the number of people inside the network. By assigning IDs to the payment card one can also capture the start and destination of each passenger.
For such an approach each entry and exit to the network must have barriers which comes with a high price. Additionally it must be assured that passengers can pass the barriers fast enough in case of an emergency which might require adjustments to the buildings. A shortcoming of such approach is that the utilization of single vehicles is not captured.
At entrances and exits with security cameras
The in and outflow into the network can also be measured utilizing security cameras. This approach can not capture the passenger's journey but is easier and faster to implement than using payment barriers.
Cameras inside the doors
Camera systems in doors are usually mounted in the middle of the door and capture the entry area from above. With the aid of modern image processing, cameras can detect whether a person enters or leaves the vehicle.
Security cameras inside vehicles
Security cameras are often already installed for safety reasons in public transport vehicles. The primary use case of the cameras is to provide evidence in crime investigations and to prevent crime. However, with modern image processing the video data can also be used to evaluate vehicle occupancy. Unlike cameras inside doors, which count inflow and outflow, the number of people is directly captured by the camera. This has the advantage that measurement errors do not accumulate over time.
3D-sensors or stereo cameras in doors
Dedicated 3D sensors or stereo cameras mounted inside the doors provide additional information about height of objects. While this additional information is supposed to increase robustness and accuracy it comes with a significant higher cost than security cameras.
The Isarsoft approach
The Isarsoft approach is to keep sensor costs low by utilizing security cameras. Isarsoft software is based on artificial intelligence which mimics the human capability to see, archiving high accuracy without using expensive 3D sensors.
Existing cameras can be used regardless of their positioning by adding an additional computational device either within the vehicle, in a local data center or by utilizing the computational capabilities of the Isarsoft cloud.
Isarsoft also works together with camera manufacturers to do the computation directly on the camera.
Besides counting people Isarsoft algorithms can capture other objects such as bicycles and trains which are relevant for public transport.
If you have questions about which approach is the most suitable for you do not hesitate to contact us. Our experienced consultants are happy to provide you with tailored information for your use-case.