Perbandingan Akurasi Pengenalan Jenis Beras dengan Algoritma Propagasi Balik pada Beberapa Resolusi Kamera

Rice is a staple that is cooked so that it becomes rice for daily consumption. The type of rice that is often used for daily consumption is white rice. There are several types of white rice circulating in the market that are consumed by the public. Each type of rice gives different scent, taste and price. This study compares the accuracy of white rice type recognition based on several camera resolutions. The types of rice used in this study are Jawa Barat rice, Jawa Timur rice, Pandan Wangi rice, Thailand rice and Vietnam rice. The camera resolution used is 5MP, 8 MP, 12 MP, 14 MP, and 16MP. The shooting distance used is ± 9 cm between the camera and the object of rice. The recognition method used is BackPropagation Artificial Neural Networks, while for feature extraction using the Gray Level Co-occurrence Matrix (GLCM) which consists of contrast, energy, homogeneity, and correlation. The highest results obtained at 12 MP camera resolution with the results of the recognition of 25 of 50 test data and the results of the calculation with confusion matrix obtained an average accuracy of 82%, precision of 55%, and recall of 50%. The results of this study can be used as a reference for research that uses objects of similar character, or further research with the same object in developing applications that are ready to use.