Identifying Less-Quality Burnt Clay Bricks using Image Processing

Abstract

Due to the economic crisis in Sri Lanka as well as the increase in the price of cement, the demand for burnt clay bricks has increased more than for cement bricks. When buying burnt clay bricks, special attention should be paid to their strength. Existing tests to measure their strength are impractical to use on a building site or at a brick-buying station. Since there is a lack of knowledge and experience in this field, buyers tend to buy less-quality burnt clay bricks. Because of this, a considerable amount of money is being wasted and low-quality bricks might reduce the quality or the strength of the building. The aim of this study was to assist building constructors, supervisors or any others who are interested in finding less-quality burnt clay bricks. First, a background study on clay types and quality factors of burnt clay bricks was done and, then equal data sets were collected to represent each clay type and to represent the features which were expected to be used in feature extraction. Those images were pre-processed and divided into two data sets as training and testing. Inception V3 image recognition model was used to develop the model and its accuracy was determined to be around 85%. With this developed model, even a person without proper knowledge about the field will be able to identify less-quality burnt clay bricks by just inserting an image. However, it is necessary to carry out further studies to enhance the accuracy and effectiveness of the developed model. Furthermore, the developed model can be improved as a mobile app for practical ease of use.