研究目的
To assess the potential of hyperspectral imaging technique for predicting the TVB-N content in Pacific white shrimp, and to compare spectral feature extraction methods including full spectra, SPA-selected wavelengths, and SAEs-extracted features for shrimp quality inspection.
研究成果
SAEs-based models, particularly SAEs-LS-SVM, provided the best prediction accuracy for TVB-N content with high R2 and RPD values, demonstrating the effectiveness of deep learning for feature extraction in hyperspectral imaging. This method offers a non-destructive, rapid approach for shrimp quality inspection, with potential applications in other seafood products.
研究不足
The study is limited to Pacific white shrimp under specific storage conditions; generalizability to other seafood or conditions may require further validation. Computational time for training SAEs is higher than for SPA, which could be optimized.
1：Experimental Design and Method Selection:
Hyperspectral imaging (900-1700 nm) was used for non-destructive prediction of TVB-N. Feature extraction methods included SPA and SAEs, with regression models (LS-SVM, PLSR, MLR) for prediction.
2：Sample Selection and Data Sources:
240 Pacific white shrimp samples were harvested, stored at 4°C for 168 hours, and divided into calibration (120 samples) and prediction (120 samples) sets. Hyperspectral images were acquired.
3：List of Experimental Equipment and Materials:
Hyperspectral imaging system with CCD camera, spectrograph, light source, motorized slider; shrimp samples; chemical analysis equipment for TVB-N measurement.
4：Experimental Procedures and Operational Workflow:
Samples were scanned using HSI system, images were corrected, regions of interest (ROI) were selected, spectra were extracted, feature extraction (SPA, SAEs) was performed, and regression models were built and evaluated.
5：Data Analysis Methods:
Statistical evaluation using R2, RMSE, RPD; software tools included MATLAB, Unscrambler, Keras, and ENVI.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容-
CCD camera
B1621M
Imperx Inc.
Acquiring hyperspectral images with spatial and spectral data
立即获取同款设备 -
Imaging spectrograph
ImSpectorN17E
Spectral Imaging Ltd.
Splitting light into spectral components for hyperspectral imaging
暂无现货
预约到货通知
-
Light source
3900-ER
Illumination Technology, Inc.
Providing illumination for hyperspectral imaging
暂无现货
预约到货通知
-
Stepper motor
IRCP0076-1COMB
Isuzu Optics Corp.
Moving the sample holder for line-scan imaging
暂无现货
预约到货通知
-
Oxygen-increasing machine
Airpump100-3704
Eheim GmbH & Co. KG
Adding oxygen to seawater for shrimp transport
暂无现货
预约到货通知
-
Optical lens
F/1.4, f=23 mm, 21e1001917
Schneider Optics Inc.
Focusing light onto the CCD camera
暂无现货
预约到货通知
-
White Teflon board
Isuzu Optics Corp.
Serving as a white reference for image correction
暂无现货
预约到货通知
-
登录查看剩余5件设备及参数对照表
查看全部
