Identification of Cement and Sand Content Through Surface Image Using Image Block Techniques

Abstract

This research raises the topic of identifying the types of cement and sand mixtures on dry material using artificial intelligence. This is done because the comparison of the mixture between cement and sand is very influential on the quality of the material produced. Several experimental models affect the level of recognition accuracy. In this study the experimental model used was the image block and LBP image techniques, with a mixture of cement and sand used was 1: 1, 1: 1.5, 1: 2, 1: 2.5, 1: 3, and 1: 3.5. The recognition method used is Artificial Neural Network (ANN) with back propagation algorithm. The number of ANN training samples is 600 samples, and 120 samples for testing. This research uses image block technique before feature extraction is carried out. The features used are the mean, standard deviation, entropy, skewness, and kurtosis of LBP images. ANN training results get a three-layer hidden architecture, with testing showing an accuracy rate of 80% recognition.