EVALUATION OF SOYBEAN PLANT SHAPE BY MULTILAYER PERCEPTRON WITH DIRECT IMAGE INPUT

Evaluation of soybean plant shape in breeding is empirical and based o n visual judgment, making it unstable and inefficient and pointing up the need for a quantitative alternative. Previous studies successfully applied evaluation by linear discriminant function, fuzzy logic or ne ural network, but these models required definition and selection of im portant features for judging shape. We developed a method based on a m ultilayer perceptron (MLP) with direct image input of binary soybean i mages which does not require any shape features. An MLP is a kind of n eural networks, and can exhibit good performance in pattern recognitio n. A neural network is composed of units being simple processors, and connections between the units carrying numeric data from one unit to a nother. Units of an MLP are arranged on the layers, and connected each other between the adjoined layers. We used 326 soybean plant images j udged either ''Good'', ''Fair'' or ''Poor'' by expert soybean breeders . The images were divided into supervisor and test data sets. We studi ed 175 different MLP structures, varying the number of layers, units a nd connections. After training each MLP with the supervisor data set, we evaluated matches between MLP output and breeder judgment with the test data set. The MLP with three layers, 8x8 input units, 16 hidden u nits and three output units proved to be the superior structure. Altho ugh performance in judgment was no higher than that of previous ones, our method has the decided advantage of not requiring definition and e xtraction of the shape features and may be applicable to other crops. We should note that the MLP structure is too complicated for us to und erstand the manner of breeders' empirical judgment through this model; that is, the MLP is almost a black box for us for the time being.