Abstract Recent studies have shown that, alongside microwave-based and short-wave infrared (SWIR) technologies, terahertz-based sensing is a promising and reliable method for non-destructive testing (NDT) of food in order to detect foreign objects. This improves the safety and quality of low-hydration processed or natural foods as it is expanding the sensing capabilities in the electromagnetic spectrum by revealing further insights thanks to the higher resolution and penetration depth. Despite this broad coverage, limitations and drawbacks still remain in terms of accuracy and performance under industrial conditions and costs, particularly in the terahertz region where penetration depth decreases significantly. In order to address this crucial point, we have developed an NDT system based on a commercial-off-the-shelf product for terahertz imaging and investigated its capabilities primarily with regard to the detection of low-density foreign objects (LDFO) of varying sizes and materials. The evaluation is based on selected image quality measurement indices, e.g. Structural Similarity Index (SSIM) and variants, Feature Similarity Index (FSIM), Chi-squared distance, Mean Opinion Score (MOS), which were applied to images of different product types, foreign objects, and levels of attenuation. To this end, we introduce and make publicly available a novel dataset of industrial confectioneries ranging from wafers to chocolate-covered cookies. In addition, our presented framework offers a range of functions for comprehensive quality assessment, including the measurement process, a streamlined approach for postprocessing sub-terahertz data, and evaluation methods. According to results of threshold-based binary classification a total accuracy of 84 % was achieved.
Papadopoulos et al. (Sat,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: