Recognition of tomato nutrient deficiency using aritificial neural network based on computer vision

The main objective of this project is to develop a recognition system for tomato nutrient deficiency based on artificial neural network to assist fertilization management during the growth of tomato. Digital images of tomato leaves, which were in deficiency of nitrogen, magnesium or Fe nutrition element, were collected from May to October in 2003 using the CCD camera. New color feature extraction methods based on average percentage Hue histogram were developed to distinguish ill leaves from normal leaves. The textural features of individual ill tomato leaves were extracted from different methods. Of the five textural features used in the discriminant analysis, two were computed from grads image, one (Inertia) was extracted from grads-gray co-occurrence, and the remaining two(Entropy, Variance) were computed from the wavelet decomposition. Using the above color and textural features as input vectors of the recognition system, the classification accuracies of the testing data set were 95%,92.5%, 92.5%, 87%, and 87％, respectively, for normal old leaf, normal new leaf, deficiency of Fe, deficiency of nitrogen, and deficiency of magnesium.