Original Source Here
Breezing Through TSA Checkpoints – AI/ML Assisted Self Screening
Imagine breezing through airport security checkpoints without interacting with TSA officers. TSA may be thinking that the feeling is mutual! In this article, I write about a program known as Screening at Scale (SaS), a proposal to implement self-service screening at TSA checkpoints, and how AI is helping.
Let’s address the burning question right away. SaS program explicitly states that any type of recognition or identification of people is not to be used in this program. In addition, you do not have to pay, you do not have to enroll/register. So, please breath easy, and continue to read.
The idea is to implement something similar the self-checkout registers at retail stores. You enter a booth, follow instructions, clear yourself and you are in! The diagram below provides a visualization of such a booth.
How does this improve the efficiency and passenger experience? To begin with, these are self service booths, and there is no TSA officer in there. In addition, this booth should provide significantly improved privacy and comfort compared to pat-downs. Finally, it is expected that the self-screening booths are compact. See the diagram below.
Role of Deep Learning and Computer Vision
Deep learning and computer vision can help in two ways:
- Operational: understand what is going on in and around the booth, and provide feedback to passengers in real time. Seek help in cases where passengers are not successful in self-screening.
- Analytics: measure KPI indicators such as how long it takes for each self screening process, and what are some of the most commonly encountered difficulties.
Several types of algorithms are needed and will find use in SaS use cases.
Object Detection Algorithms
These algorithms are trained to detect one or more types of “objects” in videos. Object detection is a well-established process in computer vision. Typically we need to collect several thousand images of each type of object to be recognized, and train a computer program to detect similar images.
SaS has several types of objects to be detected. The following is a summary:
- Person detection: detect individuals and families as they enter the booth.
- Person classification: classify people as adult, child, passengers with special needs etc. to initiate different processes variations.
- Special equipment and accommodation detection: for example, wheelchairs and strollers.
- Baggage detection: several kinds of bags need to be detected. Creating a model to identify all kinds of bags will be quite challenging, but it can be done with enough training images.
- Personal items detection: for example, shoes, jackets, hats, belts.
Action Recognition Algorithms
These algorithms are trained to detect specific types of actions in still images and full motion videos. Example actions: emptying pockets, removing shoes, removing jacket, removing belts, placing bags on the table.
Anatoly Detection Algorithms
Special-purpose algorithms need to developed in order to identify certain situations. For example, someone may be raising their hand(s) — perhaps to ask for help. Someone may be going back and forth too many times through the system, indicating some kind of difficulty following instructions. It may also be necessary to deploy such algorithms to identify special situations right at the entrance in order to route special-situation passengers to assisted screening before they enter the self service booths.
System Of Systems
A fully functional POD with multiple booths will need several non AI/ML components.
- The POD and booth hardware, with all the features needed to fully divest personal possessions and clear the screening process.
- A display screen, speaker and microphone setup to show instructions and establish bi-directional audio communication in case there is need to establish a voice connection with a remote support person / TSA officer.
- X-ray and metal scanning equipment to clear the bags and personal possessions.
- Finally, a plan to re-route the passenger to the assisted screening station.
Now that you have made it to the end of this article, you are all cleared to fly! You can find additional information about this program at DHS Site.
Trending AI/ML Article Identified & Digested via Granola by Ramsey Elbasheer; a Machine-Driven RSS Bot