While the early concept of artificial intelligence has roots as far back as 1935, it wasn’t put to practical use consistently until the 2000s. Many of us use AI every day when we unlock our phones via facial recognition, while others use the technology for text analysis and generation in the form of ChatGPT. In the public safety space, AI has been adopted to help with a range of tasks, from assisting with report writing to easing the challenge of personnel scheduling.
As the applications for AI continue to grow, those in corrections are finding the technology beneficial, particularly when it comes to safety measures. Using AI technology to assist in contraband detection doesn’t replace the need for corrections officers. Instead, it acts as an additional screening layer, helping ensure prohibited items aren’t missed and reducing the potential for human error.
Unlike what’s often depicted in movies or on television, AI doesn’t run rampant with its own thought processes, overtaking the inputs from human operators. In the case of contraband detection, it’s a specialized science that’s developed by experts in a lab who craft specific software algorithms.
UNDERSTANDING THE ROLE OF AI IN CONTRABAND SCREENING
No matter how sophisticated a correctional facility might be, there’s one constant that remains true – keeping contraband out is a top priority to keep staff and inmates safe. Various types of screening measures can be implemented to help detect prohibited items, from X-ray systems to physical pat-downs, but unfortunately, smuggling methods are becoming increasingly complex.
In general, screening for contraband follows the guideline that anything appearing on a high-resolution search larger than a human thumbnail needs to be further investigated. While this is helpful when it comes to detecting a wide variety of narcotics, weapons and other threats, it can also become labor-intensive to conduct additional searches. When a Band-Aid on a person triggers an alarm, for example, a secondary inspection may be required to make sure the bandaged area is not also hiding narcotic-soaked papers.
Because contraband, particularly weapons, can take the form of many shapes and sizes, AI used in screening systems needs to be trained to determine what is a threat and what isn’t. Unlike in aviation, where items like a toothbrush or a handful of pills may be benign from a security perspective, using artificial intelligence in a corrections setting presents a greater challenge when trying to determine what items may actually pose a concern.
HOW DO YOU TRAIN A COMPUTER?
Artificial intelligence can seem shockingly smart, but it’s still just a type of technology that needs human input to be truly effective. On a basic level, using AI in correctional facility screening systems requires teaching the program patterns it can use to determine if an object is a threat or a non-threat.
The initial stages of training are labor-intensive, with scans performed countless times while positioning an object in various locations of the body. To offer consistent detection capabilities, these scans must be completed using individuals of varying gender, height, weight and dress. Certain anomalies, like cultural dress or body-mass index outliers, also need to be included for a more complete set of training data.
The resulting database can be used to develop varying types of AI models depending on a correctional facility’s needs as well as work to evaluate current system performance and help build new software releases.
THE SCIENCE BEHIND THE SCIENCE
Conducting numerous scans to train on threat detection isn’t as simple as asking individuals to walk through a screening system, as the technology used is highly complex. Even still, machine learning used in AI applications can only go so far, as security system screening needs to extend beyond the limitations of machine learning.
One can never include every weapon known to man in every shape, so understanding what are clothes, for instance, and what are anomalies becomes an important building block for AI. This level of anomaly detection while reducing false alarm rates requires extensive AI training.
“Thinking you need to have the system recognize every single object is not a good way to go,” explained Darren McCarthy, technical marketing manager at Rohde & Schwarz. “You’re never going to get there but being able to define threat alarms into different types of object classes, like clothing for example, allows you to not have to learn every single object.”
Since Rohde & Schwarz scanners use millimeter wave technology, the systems can learn to detect objects through the way each object absorbs and reflects the waves. When an object is scanned, the data generated is displayed as pixels within an image region. Determining whether the region is normal or suspicious and considering the features of the object – surface roughness or density, for example – are what helps the AI algorithm learn when to sound an alarm.
Adding to the object library is an ongoing endeavor, as the company uses a repository of privately collected images gathered with consent from its 10,000-plus employees. This allows the AI algorithm to be trained on new objects quickly and presents opportunities to fine-tune certain anomalies like face masks or cultural dress.
A PERFECT MATCH FOR CORRECTIONS
There is no one-size-fits-all when it comes to scanning AI algorithms, as different facilities will have varying priorities. Rohde & Schwarz develops custom-tailored versions of its firmware to account for these requests, allowing facilities to specify which features are most valuable for their needs. A partnership with the Norwegian Prison Association for the last five years has further helped develop each algorithm to better serve the needs of correctional facilities in the United States and abroad.
The threat of contraband in correctional facilities is constantly evolving and requires systems that can adapt to new challenges at a moment’s notice. By integrating Rohde & Schwarz scanning systems into a facility’s screening practice, cognitive AI algorithms can help ensure contraband detection efforts remain one step ahead.
Visit Rohde & Schwarz for more information.
