研究目的
To develop a non-destructive method for quantitative determination of cadmium content in tomato leaves using hyperspectral imaging technology and feature selection algorithms to improve accuracy and efficiency.
研究成果
The hyperspectral imaging technology combined with BOSS feature selection is feasible for rapid and accurate prediction of cadmium content in tomato leaves, achieving high calibration and prediction accuracy. This method provides a non-destructive alternative to traditional techniques and has potential applications in other crops.
研究不足
The accuracy of prediction still has a gap compared to classic methods and should be further improved. The study did not compare cadmium stress with other stresses like nutrient or water stress, and differences in external and internal components were not fully explored.
1：Experimental Design and Method Selection:
The study used hyperspectral imaging to acquire spectral data from tomato leaves under different cadmium stress gradients. Feature selection algorithms (CARS, VCPA, BOSS) were applied to reduce data dimensionality, and partial least squares regression (PLSR) models were built for prediction.
2：Sample Selection and Data Sources:
Tomato plants (F1Hybird, Guang Zhou, China) were cultivated in perlite with seven cadmium stress gradients (0, 0.2, 0.5, 1, 2, 5, 10 mg/L). A total of 350 leaf samples (50 per gradient) were collected and analyzed.
3：2, 5, 1, 2, 5, 10 mg/L). A total of 350 leaf samples (50 per gradient) were collected and analyzed. List of Experimental Equipment and Materials:
3. List of Experimental Equipment and Materials: Hyperspectral imaging system (ImSpector V10E camera, Spectral Imaging Ltd; halogen lamps; control cabinet SC10; electric control displacement table MTS120; computer with Spectral-Cube software), graphite furnace atomic absorption spectrometry (GF-AAS) for cadmium content determination, and software (Matlab R2013a, Unscrambler 10.1, ENVI).
4：1, ENVI). Experimental Procedures and Operational Workflow:
4. Experimental Procedures and Operational Workflow: Hyperspectral images were acquired in a dark environment, preheated for 30 minutes, with exposure time set to 50 ms. Spectral data were extracted using region of interest (ROI) selection. Cadmium content was measured using GF-AAS. Feature selection and PLSR modeling were performed.
5：Data Analysis Methods:
Statistical analysis included calculation of R2 and RMSE for model evaluation. Feature selection algorithms were used to identify optimal wavelengths, and PLSR was employed for regression modeling.
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-
White Reference
Teflon white board
Edmund Optics Inc.
Used for image correction to obtain relative reflectance.
-
Hyperspectral Image Camera
ImSpector V10E
Spectral Imaging Ltd
Acquiring hyperspectral images of tomato leaf samples in the visible to near-infrared range (430-960 nm).
-
Halogen Lamp
Type of 2900
Illumination Technologies
Providing illumination for hyperspectral image acquisition.
-
Control Cabinet
SC10
Beijing Optical Instrument Factory
Controlling the hyperspectral imaging system.
-
Electric Control Displacement Table
MTS120
Beijing Optical Instrument Factory
Moving samples during hyperspectral image acquisition.
-
Software
Spectral-Cube
Spectral Imaging Ltd
Data acquisition for hyperspectral images.
-
Software
Matlab R2013a
The Mathworks Inc.
Data analysis and feature selection algorithms execution.
-
Software
Unscrambler 10.1
CAMO
Chemometric analysis for building regression models.
-
Software
ENVI
ITT visual information solutions
Selecting regions of interest (ROI) in hyperspectral images.
-
Graphite Furnace Atomic Absorption Spectrometry
Determining cadmium content in tomato leaf samples based on national standard GB5009.15-2014.
-
登录查看剩余8件设备及参数对照表
查看全部
