Laser & Optoelectronics Progress, Volume. 58, Issue 21, 2128001(2021)
Effect of Spectral Configuration on Soil Organic Matter and Electrical Conductivity Predicted by Optimal Band Combination Algorithm
Visible light and near infrared (Vis-NIR) spectrometer is a practical tool that can be used for alternative soil physical and chemical analysis to assess soil properties. The optimal band combination algorithm is an effective method to extract spectral variables by considering the interaction information between bands. However, for laboratory Vis-NIR spectral data, this method is vulnerable to the “dimension disaster”.A method that appropriately reduces the spectral configuration (reducing the number of spectral bands and coarsening the spectral resolution) to improve the calculation efficiency without significantly affecting the prediction accuracy is proposed. First, six spectral configurations are constructed, which means that the number of spectral bands is reduced from 2001 to 19, the spectral resolution is increased from 3 nm to 100 nm, and the spectral sampling interval is equal to the spectral resolution (uniformly spaced sampling). Then, partial least squares regression combined with the optimal band combination algorithm is used to predict the optimal spectral parameters of soil organic matter (SOM) and electrical conductivity (EC) under different spectral configurations. The results show that until the spectral resolution is 60 nm (32 wavelengths per spectrum), the optimal band combination algorithm can improve the prediction accuracy compared with the full band spectrum data. The best band combination algorithm has no significant difference in prediction accuracy under 1?20 nm spectral resolution (about 2%); SOM [determination coefficient(
Get Citation
Copy Citation Text
Yi Zhang, Jianli Ding, Zipeng Zhang, Xiangyu Ge, Jinjie Wang. Effect of Spectral Configuration on Soil Organic Matter and Electrical Conductivity Predicted by Optimal Band Combination Algorithm[J]. Laser & Optoelectronics Progress, 2021, 58(21): 2128001
Category: Remote Sensing and Sensors
Received: Mar. 2, 2021
Accepted: Mar. 8, 2021
Published Online: Nov. 1, 2021
The Author Email: Ding Jianli (watarid@xju.edu)