A new method for detecting foreign objects in agricultural products
Vladyslav ´¡²Ô»å°ù¾±¾±²¹²õ³ó±ð²Ô’s research provided a faster, cost effective way to detect foreign objects in food, improving quality control and ensuring safer products for consumers.
The agricultural industry faces a major challenge: finding foreign objects, like bones in chicken or insects in fruits, that can make food unsafe or unappealing. To solve this, PhD researcher Vladyslav Andriiashen developed a new approach. Instead of relying on costly and time-consuming methods to collect real X-ray images, he used computers to generate large sets of synthetic X-ray images. This enables food inspection systems to be designed much more quickly and efficiently. He defended his thesis on Thursday, May 15.
The challenge of detecting foreign objects
The difficulty with detecting foreign objects in food comes from the natural variety in shape and composition, which differs from manufactured products. While factory-made items follow strict designs, food products can look very different from one another.
Traditional X-ray inspection systems required large amounts of real images to train their detection systems, but obtaining these images was challenging. method, which created artificial X-ray images, simplified this process and provided a cost-effective solution to improve food safety.
Improving food safety with new inspection technology
This research is part of larger focus on using X-ray technology to help industries improve their processes. This kind of technology is important not only for agriculture but also for manufacturing and security.
A similar project at CWI, called , looks at smart solutions for farming and gardening, showing how data-driven technologies can make agriculture more efficient.
Creating X-ray images with computers
In his research, Andriiashen came up with a new way to create large sets of X-ray images. Normally, gathering real X-ray images for every possible inspection scenario is expensive and time-consuming.
Instead, ´¡²Ô»å°ù¾±¾±²¹²õ³ó±ð²Ô’s method used computers to generate synthetic (fake) X-ray images, making it much quicker and cheaper to develop food inspection systems.
Phd researcher Vladyslav Andriiashen
Using artificial data to train AI systems
A key finding from the research is that AI systems (called neural networks) could be trained using these synthetic X-ray images and then perform effectively with real X-ray images.
This worked because many of the details in real X-ray images could be simplified without losing accuracy. Even with simpler data, the system could still detect foreign objects in food, like bones or insects.
Faster, cheaper system design
One of the major benefits of this method is that it allows for the design of X-ray inspection systems without needing a lot of real world testing. his can reduce development time, save money, and improve the performance of the systems. After using synthetic data to create the design, only the final system needs to be tested in real world conditions.
Making food inspection more efficient
In conclusion, ´¡²Ô»å°ù¾±¾±²¹²õ³ó±ð²Ô’s research offers a faster and cheaper way to develop automated inspection systems. This new approach can help the agricultural industry maintain high food quality standards while keeping up with the demand for large scale food production.
-
Supervisors
Robert van Liere, Tristan van Leeuwen (external)
Written by
Latest news
