Novelty detection of foreign objects in food using multi-modal X-ray imaging. (September 2016)

• Record Type: Journal Article
• Authors: Einarsdóttir, Hildur; Emerson, Monica Jane; Clemmensen, Line Harder; Scherer, Kai; Willer, Konstantin; Bech, Martin; Larsen, Rasmus; Ersbøll, Bjarne Kjær; Pfeiffer, Franz
• Is Part Of: Food control. Volume 67(2016:Sep.)
• Page Start: 39
• Page End: 47
• Publication Date: 2016-09
• Subjects: X-ray radiography -- Dark-field imaging -- Phase-contrast imaging -- Foreign object detection -- Novelty detection -- Texture analysis; Food -- Quality -- Periodicals; Food -- Analysis -- Periodicals; Food handling -- Periodicals; Food industry and trade -- Quality control -- Periodicals; Aliments -- Industrie et commerce -- Qualité -- Contrôle -- Périodiques; Aliments -- Qualité -- Périodiques; Aliments -- Analyse -- Périodiques; Hygiène alimentaire -- Périodiques; Food -- Analysis; Food handling; Food -- Quality; Periodicals; Electronic journals; 664.07
• DOI: 10.1016/j.foodcont.2016.02.023
• Languages: English
• ISSNs: 0956-7135

Abstract

Abstract: In this paper we demonstrate a method for novelty detection of foreign objects in food products using grating-based multimodal X-ray imaging. With this imaging technique three modalities are available with pixel correspondence, enhancing organic materials such as wood chips, insects and soft plastics not detectable by conventional X-ray absorption radiography. We conduct experiments, where several food products are imaged with common foreign objects typically found in the food processing industry. To evaluate the benefit from using this multi-contrast X-ray technique over conventional X-ray absorption imaging, a novelty detection scheme based on well known image- and statistical analysis techniques is proposed. The results show that the presented method gives superior recognition results and highlights the advantage of grating-based imaging. Highlights: Grating-based X-ray imaging provides superior contrast of foreign objects in food. Organic materials are better visible than with conventional absorption X-ray imaging. Multivariate analysis combining intensity features provides higher detection rates. Objects as small as 0.5 mm can be detected.