Aplikasi Jaringan Syaraf Tiruan dan Analisis Komponen Utama untuk Sortasi Mentimun

Abstract

This paper discusses the development of a software prototype for cucumbers selection and grading by applying Standard Backpropagation Neural Network (SBPNN) and Principal Component Analysis (PCA). The prototype has been tested to recognize cucumbers based on their shapes (i.e. straight or non-staright cucumbers). Cucumbers' images data were expressed in eight position of rotational axes: 0°, 4S0, 90°, 139, 180°, 229, 270°, 315'. The implemented system can recognized 100 % of all tested straight cucumbers and 75% of all tested non-straight cucumbers. The performance implemented SBPNN was also compared to another system called Probabilistic Nural Network (PNN). The results shows that SBPNN is better than PNN in time execution; however PNN is better than SBPNN in generalization or recognition accuracy.