The Use of Hand Phone Camera to Determine Paddy Leaf Color Level as a Reference for Fertilizing Dosage

Abstract

This research aimed at attempting to use hand phone cameras to determine paddy leaf color level instead of using leaf color chart. Five brands of popular hand phones were used to capture the image of paddy leaves which were previously measured with the standard IRRI’s leaf color chart. The leaves were placed on the palm skin and on a white paper under two lighting conditions: under body shadow (around 100 lumens) and under open sunlight (around 850 lumens). The images were then processed to extract the RGB color components of both background image and the leaf image. Eighteen color components were used as the classifier. With those eighteen variables, the k-nearest neighborhood (k-NN) pattern recognition method was applied to develop the prediction tool. Almost clear distinct different of color characteristics was found between background images and leaf images, but on the other hand, the color characteristics of each level of paddy leaf color level were not distinctively different. With the k-NN prediction method it was found that the best accuracy was showed by capturing sample images on palm skin under body shadow (66%) and followed capturing sample images on white paper under open sunlight (68%). The low accuracy is probably caused by the ambiguity of leaf samples colors.