A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Data fusion modelling of visible-near-infrared and mid-infrared spectra
Spectroscopy has emerged as a solution to estimate key soil attributes in precision agriculture (PA) during recent decades. Chemometrics and machine-learning methods are used in order to extract useful information out of the spectra. In this paper, the performance of visible-near-infrared (Vis-NIR) and mid-infrared (MIR) spectrophotometers for the prediction of pH, organic carbon (OC), phosphorous (P), potassium (K), magnesium (Mg), calcium (Ca), sodium (Na), moisture content (MC), and cation exchange capacity (CEC) were evaluated. Using 267soil samples measured with a CompactSensspectrometer (tec5 technology, Germany) with 350-1700nm spectral range and a 4300-FTIR (Agilent, US) with 650-4000cm-1spectral range, we compared single-sensor partial least squares (PLS) regression after feature selection. To take advantage of both sensors, the combined use of them were evaluated in three fusion scenarios: 1. Spectral concatenation (SC) in which the raw vis-NIR and MIR spectra are concatenated; 2. Feature fusion (FF) wherein the features (i.e., selected spectral ranges) of vis-NIR and MIR are concatenated; and 3. Fusion of the predictions given by vis-NIR and MIR PLS-basedmodels by linear regression (LR). The validation results showed that the vis-NIR model outperforms the MIR model in the prediction of all studied attributes, except for pH, Ca, and CEC. Furthermore, the single-sensor accuracies were improved in all cases by LRwhile SC and FF enhanced the single-sensor accuracies just in cases of OC, Ca, and CEC with FF being superior to SC. However, the improvement achieved by fusion was not significant. Accordingly, it is suggested to use just vis-NIR for prediction of the studied soil attributes since it showed more robustness than MIR.
Data fusion modelling of visible-near-infrared and mid-infrared spectra
Spectroscopy has emerged as a solution to estimate key soil attributes in precision agriculture (PA) during recent decades. Chemometrics and machine-learning methods are used in order to extract useful information out of the spectra. In this paper, the performance of visible-near-infrared (Vis-NIR) and mid-infrared (MIR) spectrophotometers for the prediction of pH, organic carbon (OC), phosphorous (P), potassium (K), magnesium (Mg), calcium (Ca), sodium (Na), moisture content (MC), and cation exchange capacity (CEC) were evaluated. Using 267soil samples measured with a CompactSensspectrometer (tec5 technology, Germany) with 350-1700nm spectral range and a 4300-FTIR (Agilent, US) with 650-4000cm-1spectral range, we compared single-sensor partial least squares (PLS) regression after feature selection. To take advantage of both sensors, the combined use of them were evaluated in three fusion scenarios: 1. Spectral concatenation (SC) in which the raw vis-NIR and MIR spectra are concatenated; 2. Feature fusion (FF) wherein the features (i.e., selected spectral ranges) of vis-NIR and MIR are concatenated; and 3. Fusion of the predictions given by vis-NIR and MIR PLS-basedmodels by linear regression (LR). The validation results showed that the vis-NIR model outperforms the MIR model in the prediction of all studied attributes, except for pH, Ca, and CEC. Furthermore, the single-sensor accuracies were improved in all cases by LRwhile SC and FF enhanced the single-sensor accuracies just in cases of OC, Ca, and CEC with FF being superior to SC. However, the improvement achieved by fusion was not significant. Accordingly, it is suggested to use just vis-NIR for prediction of the studied soil attributes since it showed more robustness than MIR.
Data fusion modelling of visible-near-infrared and mid-infrared spectra
Javadi, Seyed Hamed (author) / Mouazen, Abdul (author)
2021-01-01
EurAgEng 2021 Conference, Proceedings
Conference paper
Electronic Resource
English
DDC:
690
Leaf Area Index Estimation from Visible and Near-Infrared Reflectance Data
| Online Contents | 1995
Leaf Area Index Estimation from Visible and Near-Infrared Reflectance Data
| Online Contents | 1995
Urban Classification by the Fusion of Thermal Infrared Hyperspectral and Visible Data
| Online Contents | 2015
Calibration of Advanced Visible and Near Infrared Radiometer (AVNIR)
| Online Contents | 1997