Skip to content

See like Superman with Matroid


Chase Hart | August 23rd, 2021


Problem: XRay Images can be hard to decipher, and even harder to verify when a checkpoint is processing hundreds of bags per hour.

The Transportation Security Administration (TSA) operates security checkpoints at airports in the United States to keep dangerous items off airplanes. At these checkpoints, the TSA employs X-ray scanners, so Transportation Security Officers (TSOs) can inspect the contents of carry-on possessions without manually digging through each object. However, identifying and locating all potential threats can be a challenging task. 

First, the set of prohibited items that TSOs must identify is quite diverse: firearms; sharp instruments; blunt weapons; and liquids, aerosols, and gels with volumes exceeding the TSA-established thresholds. 

Second, the majority of scans are standard, uninteresting, unthreatening suitcases, yet TSOs must remain alert for a potential threat.

Third, because X-ray scans are transmission images, the contents of a bag are displayed on top of each other into a single, cluttered scene, which can make identification of individual items difficult. 

Due to the high volume of images that need to be analyzed, each TSO is only allowed to look at XRay images for 30 minutes. After 30 minutes, a TSO’s visual capacity for scanning drops significantly and they are cycled to a different part of the security checkpoint pipeline. This forces a system that increases checkpoint downtime, stress on passengers and TSOs.

These scans are also dependent on the level of training each TSO has, how long they’ve been on shift, how long they’ve been looking at images, and can ultimately have a subjective bias dependent on each TSO.

If a suspicious object is missed or misjudged, it can lead to catastrophic consequences for both the safety of the travelling public and for the safety of the nation as a whole. 

Matroid’s solution: Suspicious Object Detector for XRay Images

To help solve this critical problem, Matroid has created a Suspicious Object Detector. This detector can be used as a baseline for understanding what is possible within Matroids Computer Vision Solution.

Automated inspection can be performed with a Matroid detector, an “automated pair of eyes” that leverages state-of-the-art computer vision to find objects of interest in visual media.

Using Matroid’s Studio, you can train detectors with practically any visual media dataset. For this project, we took a dataset of various different types of suspicious objects and uploaded them into Matroid Studio. After the images were uploaded, we verified the existence of suspicious objects and then trained the detector with the press of a button. Within minutes, we had a detector that could identify the presence of a suspicious object within a xray image. With less than an hour of work we had a suspicious object detector that would have otherwise taken weeks or months to train from scratch.

Matroid solution: Real time detections for better and faster inspections

Matroid can monitor for suspicious objects from XRays in real-time with Stream Monitoring. With Stream Monitoring, you can connect any visual sensor, monitor them with your detectors, and set up API and email based alerting for when objects or defects of interest are detected. For example, as a Rapiscan 620DV is scanning baggage, it will send images directly to the Matroid Stream Monitoring platform, allowing for our systems to automatically detect unwanted objects in real time. This reduces the stress on TSA agents, allowing for TSOs on site to work on other parts of the safety process, and simply verifying alarms when they occur. 

Matroid solution: Tracking suspicious objects with Similarity Search

Beyond creating the initial detector with the Matroid Studio, Matroid can also help manage suspicious object detection data with Similarity Search, a feature that enables millisecond level visual similarity search across archived or live media at scale. This is particularly useful for searching for prevalence of suspicious objects over time. It can also be used to search for unknown objects in recorded visual data, in order to be used for training examples for either the detector or for TSOs. It can also be used to track suspicious bags, persons, or objects throughout the airport after they leave the TSA Checkpoint. 

Building detectors for XRay Images is one of many examples of Matroid helping to automate visual inspection in industrial IoT, improving operational productivity, whilst minimizing costs. To learn more about how Matroid could help you, get in touch with us today.

Download Our Free

Step By Step Guide

Building Custom Computer Vision Models with Matroid

Dive into the world of personalized computer vision models with Matroid's comprehensive guide – click to download today