Crop parameter estimation of Lady finger by using different neural network training algorithms

Now a days ANN has became an important tool in crop yield prediction and crop parameter estimation. But very few reported work is available to estimate the crop parameters by using suitable ANN Training algorithms. In the present study the crop parameters of interest are soil moisture, leaf area index and biomass were estimated using ground truth microwave scatterometer data and ANN. Two training algorithms Levenberg-Marquardt (TRAINLM) and Scaled conjugate gradient training algorithm (TRAINSCG) were used to estimate the crop parameters. The estimation of parameters with minimal error obtained with the test data confirms the usefulness of our work. Training algorithm TRAINLM shows better result in comparison to TRAINSCG. This work suggests that the ANN model with training function TRAINLM and transfer function PURELIN is a promising alternative to estimate the crop parameters. The main advantage of ANN approach as estimator is that it has the potential for world wide coverage.