Pattern Recognition and Classification Using Backpropagation Neural Network Algorithm for Songket Motifs Image Retrieval

- Yuhandri, Sarifuddin Madenda, Eri Prasetyo Wibowo, - Karmilasari

Abstract


This paper proposes a method of extraction, classification and pattern recognition songket cloth texture. Features Chain-code pattern texture (Chain-code pattern texture features) are used as the basis songket search in the database or referred to as a texture-based songket pattern recognition which is part of a content-based image retrieval (CBIR). This method consists of two parts: the first is the process of establishing databases feature Chain-code pattern texture songket and training process of pattern recognition using backpropagation neural network (BPNN), the second is the retrieval process to recognize the pattern songket (songket pattern recognition and retrieval). The proposed method is a combination of several algorithms: color image segmentation, binarization, cropping, edge detection/pattern, feature extraction pattern (probability widened chain-code datasets) and BPNN training and test. Results of tests on 40 different songket motifs with training data showing the level of accuracy of the proposed method. Results of tests on 40 songket motifs show a good degree of accuracy of proposed method where the precision value reached 98% and recall value reached 99%.

Keywords


songket; moore; chain code; CBIR; BPNN; neuron; precision value; recall value

Full Text:

PDF


DOI: http://dx.doi.org/10.18517/ijaseit.7.5.2200

Refbacks

  • There are currently no refbacks.



Published by INSIGHT - Indonesian Society for Knowledge and Human Development