SISTEM PENGENALAN OBYEK REAL-TIME DENGAN JARINGAN SYARAF TIRUAN BACKPROPAGATION

This minithesis (final project) explains about object recognition system that using artificial neural network as classifier of object types. Artificial neural network (ANN) receives four object features, that is, bounding rectangle, bounding circle, first and second momen invarian of Hu. That features extracted using image processing program and good accurate level to representating object characteristics.This system developed in real-time environment. QuickCam camera used as visual sensor to capture light from system environment. System have capability to recognizing objects as continue by good accurate level is 97.14% for normal camera position. That’s mean, camera be same position when taking sample data for ANN training. System also have capability to recognizing smaller objects (a half normal size), bigger (two normal size), and shape-changed objects. For each condition above, success procentage to recognizing objects are 94.28%, 95.71, and 87.14%. The optimum structure for this ANN: four neurons for input layer, 10 neurons for hidden layer, and seven neurons for output layer with learning rate is 0.9.