Application of neural network to predict leaf area of cabbage plug seedlings population by machine vision

A three-layered neural network that contained 4 hidden units with 500 train ing times was used to model the relationship between the average leaf area, a commonly used growth index, of a cabbage plug seedlings population and t heir growth indices obtained by machine vision. The inputs to the neural ne twork were the soil coverage and the standard deviation of the lightness of pixels. The predicted leaf area of the population of test plug seedlings b ased on the neural network model fit well with the actual values and the co efficient of determination, R2. was 0.94. Copyright (C) 1998 IFAC.