Machine Learning for Cracks Level Classification on Concrete Surfaces Using Histogram Oriented Gradient Features

Early inspection of concrete specimens before the concrete structure is used is a prevention of concrete damage. Concrete damage to buildings and other infrastructures will threaten the lives of residents and add large costs to repair the damage. This research develops a model to classify the type of cracks due to damage on concrete surfaces through Image Processing technique. Furthermore, using the HOG feature extraction method uses cell size 50 60 as an extractor of numerical features and KNN as a model method in classifying images. The results of this study showed that the best model based on the accuracy of training and performance matrix in testing was Cosine KNN showing accuracy reaching 79.60%.