Fig. 1. Low-chromatic aberration foreign matters and soy protein meat samples containing foreign matters

Fig. 2. (a) Hyperspectral imaging system; (b) Hyperspectral 3D data
1: Computer; 2: Height adjustment lever; 3: Spectrometer; 4: Lens; 5: Dark box; 6: Stage; 7: Electronically controlled mobile platform; 8: Light source

Fig. 3. Detection flow chart of low-chromatic aberration foreign matters in soy protein meat

Fig. 4. Spectral/image features of artificial meat slices with foreign matters
(a): Color images of soy protein meat with foreign matters; (b): R gray images of color images; (c): G gray images of color images; (d): B gray images of color images; (e): average spectral data of soy protein meat and foreign matters

Fig. 5. Scores of the first three principal components of principal component analysis

Fig. 6. Screening results of characteristic wavelengths
(a): Root mean square error curve; (b): Schematic diagram of characteristic wavelengths

Fig. 7. Hyperspectral imaging technology and computer vision technology for foreign matter detection
(a): Artificial meat slices mixed with foreign matters; (b): Areas of interest for artificial meat slices; (c): Gray image data of the region of interest; (d): Hyperspectral image data of the region of interest; (e): Binary image of foreign object segmentation by conventional computer vision; (f): binary image after image segmentation using the best foreign object recognition model

Table 1. LDA classification results of different spectral preprocessing methods

Table 2. Foreign matters recognition rate of different pattern recognition models (%)

Table 3. BP-ANN verification results of artificial meat containing foreign matters