Fresh Concrete Image Data Set Development Using Data Augmentation Algorithm as Building Concrete Compression Identification Reference

Abstract

Image is one of the most important multimedia data that is widely used in information technology. Information can be represented by image and human eye can analyze and interpret the information in accordance with the objective. In the field of civil engineering, image processing is used to help identify and classify concrete quality classes based on a collection of concrete image data as in this study, namely classifying the concrete compressive strength in 3 qualities fc 22.83 Mpa, fc 24.90 Mpa and fc 26.98 Mpa. The purpose of this study is to develop a concrete image data set with data augmentation because in image processing, an adequate data set is needed so that machine learning can do learning on training data and model the data set perfectly. The method used in developing this data set uses augmentation data, which is a series of techniques or methods that increase the size and higher quality of the training data so that algorithm modeling using Neural Networks can be built better than 45 concrete images with 3 concrete quality classes. The types of data augmentation used are color augmentation with Contrast and position augmentation with re-scale, rotate, translation and flip. The results of each augmentation process produce 90 images of each concrete quality so that the total data set obtained is 270 concrete images. This data set will be used to build a compressive strength of concrete identification algorithm using GLRLM and obtain a compressive strength of concrete classification algorithm using a Neural Network.