Technology has reinforced all the aspects of our lives in terms of how we work, learn, travel or shop. It’s inescapable that the technological devices and services have begun to reflect the public ambitions and desires to connect with others and contribute to the world. Technology plays a major role in ensuring our safety and security and a large part of it caters the travel and transportation industry. The security screening technologies are crucial for this industry and are required to pass a number of criteria. At the least, they should ensure satisfactory level of accuracy in object recognition, identification and classification, and they should be capable of quick operation to check objects at places like airports where passenger congestion is generally high.
The traditional technology used in security scanning has several challenges to be considered and reworked upon. As per the researchers, the traditional technology of baggage scanners is based on x-ray attenuation and X-Ray refraction. Attenuation implies that the detection of threat objects depends on how different objects attenuate the X-rays. This capability is enhanced in the past by using dual-energy X-ray scanners, that use different X-Ray spectra and determine X-Ray reduction characteristics of the object. Refraction causes deviations at minute angles at the boundaries of objects. The comprehensive baggage scanner technology is based on detecting such deviations and creating images of contours of the objects scanned. This offers deep insight into contextualizing the nature of the scanned objects thus offering better detection rate and reduced false alarms.
Although this conventional technology is easily scalable to larger fields of view, the machines are smaller and less powerful. The increasing traffic volume at security areas is another major challenge in security scanning due to the rise in population and globalization. With large crowds moving to and fro, the scanning systems face bottlenecks, items are pushed into the scanner and a smooth security check process becomes hard to accomplish. Manual errors in such scanning process are unavoidable. However, problem of missing physical distinctiveness of objects through their image can be mitigated through operational adjustment. Time constraint is another shortcoming as delays in the check posts serving with manual baggage scanners result in flight delays globally which hit billions of dollars in terms of cost. All these factors collectively hinder the efficiency of the security infrastructure administration all over the world.
Artificial Intelligence is the most trending technology overcoming all the challenges discussed so far. Machine Learning and Deep Learning are widely utilized in various computer chromatography, vision and imagery applications. AI can play behind the scenes role in safety of passengers through applications in baggage scanning.
AI and ML are iteratively integrated to improve safety, cost and travel convenience to the public. CV methods are used to improve safety of travel and efficiency of screening process. ML is successfully applied to anomaly detection and binary classification thus showing instances that are not expected. Dangerous checked baggage can be identified as an anomaly-detection problem. It is challenging to recognize the space of all possible dangerous baggage configurations and this is where ML enters as a problem solver. ML is leveraged for baggage screening by using X-ray data available form existing scanning capabilities. ML-enabled algorithm can be taught to detect various types of threats in checked bag through X-ray scanning. The algorithm finds actual threats with very high probability with minuscule false alarms. The operations of travel security administration creates large amount of data on daily basis including density and atomic number of objects in the bag. ML technology is trained to reduce false positive detection with labeled data. Simulated threats are generated for ML training that can be safely scanned through the machines and tagged as threats.
The past decade has seen dramatic improvements in ability of AI-ML to solve complex problems in various domains. A tailored ML model is a promising solution to the challenges in baggage scanning domain. The large amounts of data is readily available, which can be structured and organized to train the machines using improved computational resources and detect baggage threats- in some cases, better than humans. The obvious promise is the potential to address many challenging tasks more effectively and efficiently than can be done through human perception and cognition.