Smartphone detection of minced beef adulteration

• DOI: 10.1016/j.microc.2021.106088
• Publication Date: 2021-02-17T23:43:32Z
• Authors: Weiran Song; Yong-Huan Yun; Hui Wang; Zongyu Hou; Zhe Wang
• Keywords: Sample (material); Smartphone application

Abstract This paper presents a study on detecting minced beef adulteration based on smartphone videos recorded under a sequence of varying colours. Minced beef samples were mixed with minced pork in the range of 10–100% (w/w) at 10% increments. Light with varying colours was generated on smartphone screen and used to illuminate the sample surface. Short videos were recorded by front camera and converted into spectrum-like data by image processing. Data samples were collected under different conditions in terms of type of smartphone, recording, distance and lighting condition, resulting in seven sets of data. A partial least squares regression model was used to predict the level of adulteration, yielding determination coefficients of 0.73–0.98 and the root-mean-square errors of 0.04–0.16 for prediction. Furthermore, smartphone videos were used to present distribution maps of adulteration levels. The results indicate the potential of the simple and low-cost approach in detecting adulteration of minced meat.