Deteksi Varietas Padi Menggunakan Gelombang Near Infrared dan Model Jaringan Saraf Tiruan

Abstract

Rice seeds is one of the important components in increasing rice production. Nowdays, there are many varieties that had been created. Each variety has the advantage based on its  assembly destination. A large number of varieties can cause errors in seed processing, storage and distribution, because the physical shape and size are almost similar and the appearance is very difficult to distinguish. The alternative to detect rice seed varieties are using near infrared (NIR) as sensors and artificial neural network (ANN) as a data processor. This research aims to study the NIR spectroscopy and ANN for detecting varieties of rice seed. NIR reflectance (1000-2500 nm) of 12 varieties were given pretreatment data such as first derivative, second derivative, normalization and standard normal variate. The pretreatment data ware used as input in ANN models. Each variety was consist of 12 samples and the weight of each sample was 40 grams. ANN model used was backpropagation multilayer perceptron with three layers as input, hiden, and output. Network weights ware estimated using gradient descent algorithm. The study showed that the form of NIR spectra was similar among varieties but had different absorption in intensity, so it could be used for determining the rice seed varieties. The best model was an ANN with standard normal variate pretreatment as input data.  The accuracy varieties prediction was 100% for traning, 99.1% for testing and 98.1% for validation. This research showed that the  NIR spectra and ANN model can be used for detection methods in rice varieties.