Image analysis associated with a fuzzy neural network and estimation of shoot length of regenerated rice callus

Regenerated rice callus with a shoot more than 2 cm in length could be transferred from a growth medium supplemented with sucrose as a carbon source to a medium without sucrose for acclimatization. Image analysis was applied for the automatic selection of regenerated rice callus for acclimatization. When the RGB values for each pixel were input into a fuzzy neural network (FNN), the shoot, callus, and medium regions were identified. The identification correctness of the FNN model was 95%. Using the model, a trinary image was reconstructed and thinning and the extraction of the longest path were performed for 25 images in order to predict the shoot length. The predicted lengths coincided well with the actual shoot lengths, the average error being only 1.3 mm.