Читать книгу The Digital Agricultural Revolution - Группа авторов - Страница 49

K. Krupavathi1*, M. Raghu Babu2 and A. Mani3

1Department of Irrigation and Drainage Engineering, Dr. NTR College of Agricultural Engineering, Bapatla, ANGRAU, India

2Department of Irrigation and Drainage Engineering, College of Agricultural Engineering, Madakasira, ANGRAU, India

3Department of Soil and Water Engineering, Dr. NTR College of Agricultural Engineering, Bapatla, ANGRAU, India

Abstract

Climate change causing extreme temperature events, erratic pattern of rainfall, droughts and floods poses serious limitations on agriculture, in turn requires regular crop monitoring and management of resources to get maximum yields. Food chain of the crops can be transformed by technological innovations, like mechanization, artificial intelligence and robotics, UAVs, sensors, Internet of Things (IoT), remote sensing, machine learning and deep learning in agriculture. The present study focused on ability of machine learning algorithm in integration with remote sensing in crop yield prediction of paddy and sugarcane crops at regional level. Crop-sensitive parameters extracted from high resolution LANDSAT 8 OLI imageries are used as neural network model inputs. The Feed Forward Back Propagating Neural Network (FFBPNN) models for crop yield were developed and calibrated in MATLAB environment. During training, the model perceptron’s were trained with 75 of the 100 inputs upto10,000 epochs with 1 to 10 hidden neurons. Four performance indices (coefficient of multiple determination, R²; MAE; RMSE and the average ratio between estimated yield to target crop yield [R_ratio]) were calculated to achieve optimum neural network. Accurate and stable results observed from the model for paddy with highest mean relative error as 6.166% and the lowest relative error as −0.133%. The range of R² values were 0.946 and 0.967 for training and same for testing was 0.936 and 0.950 for paddy in Kharif and Rabi seasons, whereas for sugarcane the values are 0.916 and 0.924 during testing and training, respectively. The highest MAE was 0.178 for Paddy (Rabi). The R_ratio values showed the under crop yield estimation of sugarcane crop. The model’s best performance was observed at [i+1] and [i+2] hidden nodes. The statistical analysis revealed that the reliability of the model in paddy yield estimation. However, slight under estimation of yield of the sugarcane crop indicates sensitivity of yield algorithms to crop input parameters. The results demonstrated the high efficacy of using remote sensing images and NN models to generate accurate crop yield maps and also revealed significant superiority of neural network models over conventional methods.

Keywords: Crop yield, remote sensing, neural networks, feed forward and back propagation, NDVI, APAR, crop water stress