KPM Analytics has introduced an upgraded version of its SiftAI foreign-material detection platform, adding new artificial intelligence models, a redesigned hygienic enclosure and enhanced lighting intended to improve the inspection of food products with variable colours and textures.

The SiftAI FM HD system builds on a platform first introduced in 2022 and previously deployed primarily in meat and poultry processing. According to the company, the new generation extends its applications to fresh produce, individually quick-frozen foods, dairy products and other food categories.

The system uses machine vision and AI models to identify visible foreign materials passing through a production line. These can include low-density contaminants such as plastic, rubber, wood, cardboard and fragments of personal protective equipment that may be difficult to identify using metal detectors or X-ray systems.

One of the principal additions is a model developed to detect clear plastic. KPM Analytics said the platform’s detection models have been refined using data gathered from installations operating under commercial processing conditions, with the aim of detecting more low-density materials while reducing false rejections.

“Over the past few years working alongside our key SiftAI FM users in the meat and poultry industry, we together identified clear areas where we could improve our system to broaden the product for more customers,” said Jon Gilchrist, product manager of protein and root crop vision inspection technologies at KPM Analytics.

The FM HD generation also introduces an IP69K-rated enclosure made from food-grade stainless steel and hard-coated polycarbonate. The design is intended for full-washdown production environments and, according to KPM Analytics, can be incorporated without changing a processor’s existing sanitation procedures.

A redesigned lighting system seeks to provide the inspection engine with more consistent images while limiting interference caused by reflective, wet or naturally variable products. Improved illumination is intended to help the AI distinguish contaminants from variations in the appearance and texture of the food itself.

KPM Analytics lists potatoes, fruit and vegetables, and IQF ingredients including diced potatoes among the applications supported by the wider SiftAI FM platform. The system can be integrated above existing conveyors and connected to either existing or purpose-designed rejection equipment.

Software, security and AI-model updates are maintained remotely by the company’s specialists. Models can also be trained and adjusted for individual products and processing conditions.

“Food safety is a complicated problem, and our approach for helping food companies isn’t static,” Gilchrist said. “Every SiftAI FM HD deployment gets smarter over time.”