Bayesian Optimization with Deep Learning based Crop Type Classification on UAV Imagery

Precise projections of seasonal agricultural outputs are indispensable for optimizing security of the food. But the gathering of agricultural data via seasonal agricultural studies was frequently not timely sufficient to notify private and public stakeholders regarding crop conditions at the time of growing season. Getting accurate and timely crop approximations are mainly difficult in countries with predominate smallholder farms due to the high diversity of crop types, larger amount of small plots, and intense intercropping. This study emphases on the advancement of Bayesian optimization with deep learning driven crop type classification (BODLD-CTC) technique. The presented BODLD-CTC technique examines the UAV images for the discrimination of crop types. To attain this, the presented BODLD-CTC technique applies Xception model as feature extractor. For classification purposes, the long short term memory (LSTM) model is exploited. At last, the BO algorithm is used to optimally adjust the LSTM hyperparameters and also considerably boost the classification efficiency. To demonstrate the improved outcomes of the BODLD-CTC method, a wide range of simulations were performed. Extensive comparative inspection stated the improvements of the BODLD-CTC method compared to recent models.