Please use this identifier to cite or link to this item:
Full metadata record
|dc.description.abstract||© 2017 IEEE. In civil construction industry, different types of crushed stone are used as aggregate materials. As the prices of crushed stone depend on their types, the automated system that can examine their type is needed to avoid human mistakes. This study aims to propose a novel method for classifying 5 different classes of crushed-stone images in the dump-body of a truck. Remarkably, 4 classes are defined according to 4 types of crushed stone and the other class is the empty dump-body of a truck. We create a crushed-stone predictor called CSDeep based on a convolution neural network (CNN) and the generic texture-features such as Gabor wavelet, Haralick and Laws. A CNN is a backpropagation neural network with an effective image processing tool, i.e., convolutions. The generic texture features are used to provide additional information that is missed by CNN. The set of 2,500 and 500 images equally sampled from each class are used as training and test data, respectively. The optimal set of generic texture features are chosen by an inheritable biobjective combinatorial genetic algorithm. The proposed CSDeep achieves 89.00% of test accuracy. To the best of our knowledge, CSDeep is the first predictor for crushed-stone images taken by a digital camera. The results show that the combination of generic texture-features and CNN is suggested to enhance the performance of a deep learning model.||en_US|
|dc.title||CSDeep: A crushed stone image predictor based on deep learning and intelligently selected features||en_US|
|article.title.sourcetitle||Proceeding of 2017 2nd International Conference on Information Technology, INCIT 2017||en_US|
|article.stream.affiliations||Chiang Mai University||en_US|
|article.stream.affiliations||Big Data Agency||en_US|
|Appears in Collections:||CMUL: Journal Articles|
Files in This Item:
There are no files associated with this item.
Items in CMUIR are protected by copyright, with all rights reserved, unless otherwise indicated.