| Popis: |
Heavy metal contamination in agricultural soil is currently a global issue. The traditional approaches for soil heavy metal (HM) estimation are insufficient for large-scale and in-time monitoring and assessment. AVIRIS hyperspectral imaging can be utilized in better way to estimate HM concentrations in soil. The authors employed transfer learning model to classify the images, further HM concentration estimation was compared with the actual values. Experimental findings show VGG19 outperformed other deep learning and machine learning models and yielded a consistent accuracy of 81.25% starting from epoch 134 to 200 epochs. The root means square error (RMSE) values of different heavy metals, arsenic (As), cadmium (Cd) and lead (Pb) were found to be 2.89, 0.12, and 0.22 and the mean square value (MSE) value was evaluated to be 0.96, 0.01, and 0.04, respectively. The results of HM estimation proves that the proposed technique is efficient and effective. |
