Abstract:
The people of the South Kalimantan wetland area have handmade wicker purun mats with a distinctive wicker pattern. Based on the results of the field survey, there are 13 patterns of wicker, namely: mata punai, saluang mudik, biji cangkeh, tapak catur, salapar biji waluh, mata gergaji, pelupuh, pemudang, balang batapak, ramak sahang, beleres, baramak and balang bagapit. The wealth of this craft treasure needs to be identified and documented so that it becomes a cultural icon of the people of South Kalimantan. GLCM (Gray Level Co-Occurrence Matrices) is a method for extracting digital image texture features, while ANN (Artificial Neural Network) is a method for classifying digital images. This study identified the pattern of purun mats using GLCM and ANN. The identification carried out is the exclusion of basic wicker patterns and mixed wicker patterns. GLCM is used to extract wicker texture features and ANN is used to classify these texture features. The output of the GLCM method is the value of ASM (Angular Second Moment) or energy, contrast,
correlation and IDM (Inverse Difference Moment). The four values are used as distinctive features of an image, and are used as input to ANN. ANN training is carried out to obtain the ANN model (weight and bias). The ANN model is used for testing dataset. This dataset uses 65 wicker samples consisting of 2 classes, namely: basic wicker patterns and mixed
wicker patterns. The sample data is grouped into two parts: 39 samples are used as training data, and 26 samples are used as testing data. The test results are in the form of accuracy in the recognition of purun mats patterns. The results of this study indicate that identification of purun mats produces 100% accuracy when training and 80.77% when testing. Based on the results of these tests, the GLCM and ANN methods can be used to classify digital image texture features of purun mats effectively.
