Rapid Testing of Aflatoxin by Using Image Processing and Artificial Neural Network

Abstract

Abstract Aflatoxin can be recognized clearly by using UV-light. This information is very important to develop the device for detecting the aflatoxin inside the corn by using image processing. Current research related to identification of aflatoxin has been conducted manually by the experts. This method have some weakness including subjectivity factors, inconsistent result, and time required used. Based on the problems above, it needed to create the rapid testing device for identification of aflatoxin with consistence result, accurate and easy tooperate. The research aimed to develop the device for rapid testing of aflatoxin. The method used image processing and artificial neural network. The raw material used was the corn. The image of aflatoxin taken by using digital camera (Gopro 4) and processed by image processing program. ANN model was developed with 10 input parameters, 20 hidden layers and 4 targets. The fourth of targets above were the size of aflatoxin such as 1 - 2 ppb, 2 – 3 ppb, 3 – 4 ppb dan 4 - 5 ppb. The result showed that the characteristics of image were very specific among the input parameters and the most influential to recognize the object was the longest diameter of aflatoxin. The result of training showed that the size of aflatoxin can be recognized by the system was 100%, while the validation by using the different sample was 99%. Based on this research can be conclude that image processing can be used as the rapid testing of aflatoxin.