EVALUATION OF SOIL FERTILITY STATUS THROUGH MIXED PIXELS ANALYSIS OF HYPERSPECTRAL DATA

Abstract

Fertility status as well as management of nutrients of soil are important for quality as well as quantity of crop yield. The appropriate nutrient type and quantity increase the crop quality and yield, on the other hand insufficient and unreasonable nutrients create a problem like increased input costs, unhealthy crops, poor yield, and environmental pollution. Thus, information about fertility status of soil is required for application of appropriate amount of nutrients in agricultural field. Evaluation of soil fertility status by traditional laboratory based chemical analysis methods are time-consuming, laborious, destructive and highly cost-intensive. To overcome these issues, hyperspectral imaging (HSI) technique is getting employed for identification and determination of soil salinity status and micronutrients in soils. HSI technique is non-destructive in nature and do not cause any pollution. However, mixed pixel hyperspectral analysis of soil for evaluation of its salt and micronutrient contents has not been done so for. So, the objective of this research work is to assess soil fertility status through mixed pixels’ analysis of hyperspectral data. The study intends to introduce a method of integration of spectral derivative and linear mixing model, coined as derivative analysis for spectral unmixing (DASU) followed by deep learning (DL) technique, a novel approach. Spectral derivative and continuum removal techniques were used for the spectral features analysis of the study samples. In this research, the estimation of fractional abundances of three component materials (salt minerals, NPK compost, and urea fertilizer) mixed with soils (sandy, loamy, and silt clay) have been carried out independently. Soil samples for the study were collected from areas located in Roorkee, Uttarakhand, India. Three types of mixed soil samples were prepared in the laboratory. The spectral reflectance curve (SRC), of the study samples acquired by a spectroradiometer in the laboratory, were used for assessment of contents of salinity and micronutrients by applying the proposed methods.