Desain Sistem Pendeteksi untuk Citra Base Sub-assembly dengan Algoritma Backpropagation
Abstract: Object
identification technique using machine vision has been implemented in
industrial of electronic manufacturers for years. This technique is commonly
used for reject detection (for disqualified product based on existing standard)
or defect detection. This research aims to build a reject detector of
sub-assembly condition which is differed by two conditions that are missing screw
and wrong position screw using neural network backpropagation. The image taken
using camera will be converted into grayscale before it is processed in
backpropagation methods to generate a weight value. The experiment result shows
that the network architecture with two layers has the most excellent accuracy
level. Using learning rate of 0.5, target error 0.015%, and the number of node
1 of 100 and node 2 of 50, the successive rate for sub-assembly detection in
right condition reached 99.02% while no error occurs in detecting the wrong
condition of Sub-assembly (missing screw and wrong position screw).
Penulis: Kasdianto, Siti Aisyah
Kode Jurnal: jptlisetrodd170365
