The scaling rate of carp is one of the most important factors restricting the automation and intelligence level of carp processing. In order to solve the shortcomings of the commonly-used manual detection, this paper aimed to study the potential of hyperspectral technology (400–1024.7 nm) in detecting the scaling rate of carp. The whole fish body was divided into three regions (belly, back, and tail) for analysis because spectral responses are different for different regions. Different preprocessing methods, including Savitzky–Golay (SG), first derivative (FD), multivariate scattering correction (MSC), and standard normal variate (SNV) were applied for spectrum pretreatment. Then, the successive projections algorithm (SPA), regression coefficient (RC), and two-dimensional correlation spectroscopy (2D-COS) were applied for selecting characteristic wavelengths (CWs), respectively. The partial least square regression (PLSR) models for scaling rate detection using full wavelengths (FWs) and CWs were established. According to the modeling results, FD-RC-PLSR, SNV-SPA-PLSR, and SNV-RC-PLSR were determined to be the optimal models for predicting the scaling rate in the back (the coefficient of determination in calibration set (R_C²) = 96.23%, the coefficient of determination in prediction set (R_P²) = 95.55%, root mean square error by calibration (RMSEC) = 6.20%, the root mean square error by prediction (RMSEP)= 7.54%, and the relative percent deviation (RPD) = 3.98), belly (R_C² = 93.44%, R_P² = 90.81%, RMSEC = 8.05%, RMSEP = 9.13%, and RPD = 3.07) and tail (R_C² = 95.34%, R_P² = 93.71%, RMSEC = 6.66%, RMSEP = 8.37%, and RPD = 3.42) regions, respectively. It can be seen that PLSR integrated with specific pretreatment and dimension reduction methods had great potential for scaling rate detection in different carp regions. These results confirmed the possibility of using hyperspectral technology in nondestructive and convenient detection of the scaling rate of carp.

With the development of optical and spectroscopy sensors, hyperspectral technology has been wildly used in the quality monitoring of food processing. Drying is one of the most commonly used processes in food processing. Moisture content (MC) and moisture distribution (MD) are very important in evaluating a drying technique and the quality of dried final products. This study aimed to investigate the feasibility of using hyperspectral data (387.1–1024.7 nm) to predict MC and realize visualization of MD during constant and intermittent drying, leading to determine the optimal drying in a nondestructive and convenient way. Two characteristic wavelength (CW) selected methods were established: regression coefficients (RC) and two-dimensional correlation spectroscopy (2 D-COS), resulting in 4 (405.7, 513.2, 606.9 and 967.1 nm) and 3 (402.3, 511.5 and 965.2 nm) CWs, respectively. The partial least-square regression (PLSR) models (FW-PLSR, RC-PLSR, and 2 D-COS-PLSR) for MC prediction were built using FW and CW. Comparing the performance of the models, RC-PLSR model was determined to be optimal (variable number = 4, R_C²=0.9731, R_CV² = 0.9703, R_P²=0.9662, RMSEC = 0.0335, RMSECV = 0.0355, RMSEP = 0.0353, and RPD = 5.5198). The visualization maps for MC and MD were generated by the optimized model, which revealed that the intermittent drying process of drying for 4 h with 1 h tempering could improve the product quality of scallops. This study proposed the method for predicting MC and MD and highlighted the potential of hyperspectral technology in the monitoring of constant and intermittent drying in scallop.

Fish head cutting is one of the most important processes during fish pre-processing. At present, the identification of cutting positions mainly depends on manual experience, which cannot meet the requirements of large-scale production lines. In this paper, a fast and contactless identification method of cutting position was carried out by using a constructed line laser data acquisition system. The fish surface data were collected by a linear laser scanning sensor, and Principal Component Analysis (PCA) was used to reduce the dimensions of the dorsal and abdominal boundary data. Based on the dimension data, Least Squares Support Vector Machines (LS-SVMs), Particle Swarm Optimization-Back Propagation (PSO-BP) networks, and Long and Short Term Memory (LSTM) neural networks were applied for fish head cutting position identification model establishment. According to the results, the LSTM model was considered to be the best prediction model with a determination coefficient (${\mathrm{R}}^2$) value, root mean square error ($\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}$), mean absolute error ($\mathrm{M}\mathrm{A}\mathrm{E}$), and residual predictive deviation ($\mathrm{R}\mathrm{P}\mathrm{D}$) of 0.9480, 0.2957, 0.1933, and 3.1426, respectively. This study demonstrated the reliability of combining line laser scanning techniques with machine learning using LSTM to identify the fish head cutting position accurately and quickly. It can provide a theoretical reference for the development of intelligent processing and intelligent cutting equipment for fish.