Soil organic matter is closely associated with soil fertility and plant growth, but conventional laboratory analysis can be slow and dependent on chemical processing. A peer-reviewed study published in Sensors investigated whether hyperspectral information could be combined with image color and texture features to improve quantitative prediction of organic matter in tea garden soil.
The researchers used an iSpecHyper-VS1000-Lab laboratory hyperspectral imaging system to acquire images from 150 soil samples. The published workflow connects sample preparation, calibrated hyperspectral acquisition, feature selection, image analysis and regression modeling in one traceable experiment.
Paper and Instrument Details
- Paper: Research on the Detection Method of Organic Matter in Tea Garden Soil Based on Image Information and Hyperspectral Data Fusion
- Journal: Sensors
- Published: 7 December 2023
- Volume and issue: 23(24), Article 9684
- First author: Haowen Zhang (the paper marks Haowen Zhang and Qinghai He as equal contributors)
- DOI: 10.3390/s23249684
- Hyperspectral system: iSpecHyper-VS1000-Lab
Study Design and Hyperspectral Acquisition
The study collected 150 soil samples from representative tea gardens in three locations in Shandong Province. Samples were prepared for laboratory measurement, and their organic matter content was determined using the applicable reference method before model development.
For hyperspectral acquisition, the paper describes a laboratory system with an integrated dark box, hyperspectral camera, simulated sunlight source, diffuse reflectance references and a linear sample stage. The stated acquisition range was 300–1000 nm with a spectral resolution of 2.5 nm. The equipment was warmed up before measurement, and dark-current and reference data were collected for calibration.
Combining Spectral, Color and Texture Information
Ten regions of interest were selected from each hyperspectral image. The researchers extracted spectral information together with nine color features and five texture features. Spectral preprocessing included standard normal variate, multisource scattering correction and smoothing. Random frog, VCPA and VCPA-IRIV methods were evaluated for characteristic-band selection.
The study then compared partial least squares regression and support vector regression models. According to the published results, the strongest reported combination was MSC + VCPA-IRIV + SVR, with R²C of 0.995, R²P of 0.986 and RPD of 8.155. These figures describe this dataset and experimental method rather than a universal instrument guarantee.
Manufacturer Name in the Published Paper
The methods section names the equipment as “ISpecHyper-VS1000-Lab” and spells the supplier as “Lyson Optics.” The model designation corresponds to the LiSen Optics iSpecHyper-VS1000-Lab platform. Readers can review the publisher page and the freely accessible full text to verify the wording and experimental configuration directly.
Why This Citation Matters
The paper provides a detailed, reproducible example of hyperspectral image acquisition being combined with machine-vision features for agricultural soil analysis. It also shows why calibrated acquisition and data fusion can be valuable when a single information source does not fully describe a complex sample.
Read the Paper on Sensors View the Free Full Text on PMC
Editorial note: This article summarizes the cited publication. The experimental design, analysis and conclusions are those of the paper’s authors. Reported results have not been independently reproduced by LiSen Optics unless explicitly stated.