Optimised multitask learning model-based deep CNN for digital soil map generation with Corvis-echo optimisation

Estimation of precise and cost-effective mapping of soil textures is crucial to track the soil properties and geographical distribution of heavy metals in soil for sustainable soil utilisation. Conventional Digital Soil Mapping (DSM) algorithms face difficulty in modelling the soil spatial variation resulting in suboptimal performance. Hence, a novel Corvis-Echo Optimised Multitasks Learning-based Deep Convolution Neural Network (CEO-based MTL Deep CNN) is proposed for generating DSMs effectively. In the proposed research, the MTL-based Deep CNN model is exploited to capture the local patterns and spatial relationships from the input data and facilitates simultaneous modelling of various soil properties. Specifically, the Corvis-Echo Optimisation (CEO) is applied for fine-tuning the hyperparameters of the classifier to attain better performance. The experimental results demonstrate that the proposed model attains the accuracy, precision, recall, F1-measure, and delay of 98.81%, 99.55%, 98.75%, 99.15%, and 317.61 ms respectively by varying the epochs 100.