Intelligent image analysis (IIA) using artificial neural network (ANN) for non-invasive estimation of chlorophyll content in micropropagated plants of potato

An artificial neural network (ANN) was used to analyze photometric features extracted from the digitized images of leaves from in vitro-regenerated potato plants for non-invasive estimation of chlorophyll content. A MATLABA (R)-based, feed-forward, backpropagation-type network was developed for an input layer (three input elements), with one hidden layer (one node) and one output layer representing the predicted chlorophyll content. A significant influence of training function during optimization of ANN modeling was observed. Among the 11 training functions tested, "trainlm" was found to be the best on the basis of comparative analysis of root-mean-square error (RMSE) at zero epoch. A significant correlation between the model-predicted and Soil-Plant Analysis Development (SPAD) meter-measured relative chlorophyll contents was obtained when the mean brightness ratio (rgb) parameters were used. Compared to a red (R), green (G), and blue (B) color space model, the rgb model exhibited better performance with a significant correlation (R (2) = 0.85). Incorporation of photometric features, such as luminosity (L), blue (B)/L, and green (G)/L, with rgb failed to improve the performance of the network. The developed Intelligent image analysis (IIA) system was able to estimate in real time the chlorophyll content of in vitro-regenerated leaves for assessment of plant nutrient status during micropropagation.