HYBRID DEEP LEARNING WITH MULTIMODAL DATA AUGMENTATION FOR RIVERBANK TREE RING CLASSIFICATION

The purpose of this research was to develop a model for counting or classifying annual tree rings lamphu trees along riverbanks. study based on image processing combined with deep learning and machine the model. data augmentations were processed through preprocessing region interest, denoising, color conversion, blurring, thresholding, dilation. This process transforms one image’s into seventy images as dataset. Deep models developed Visual Geometry Group-16-based convolutional neural networks applied different support vector classifiers: L2 norm, categorical hinge loss, Weston-Watkins. results showed that loss-support classifier gave highest efficiency an accuracy 94.07% cross-entropy loss 5.48%. Weston-Watkins norm followed by 93.70% 93.26%, respectively. Softmax function lowest order 92.96%. Keywords: Convolutional Neural Network, Data Augmentation, Learning, Support Vector Machine, Tree Ring DOI： https://doi.org/10.35741/issn.0258-2724.58.3.27