Development of Automated Web-Based Condition Survey System for Heritage Monuments Using Deep Learning

Abstract

AbstractHeritage is the identity of any society, and it should be conserved for the coming generations. Heritage monuments and buildings are seriously threatened by environmental agencies such as moisture, intense solar radiation, prevailing winds, and rain, changing their physical attributes. These environmental agencies cause various defects like spalling, abrasion, cracks, stains, and fungal growth in the heritage monuments. Detection and documentation of these defects are processes of the condition survey, which should be required just before the commencement of repair and maintenance of any structure. Traditionally, these defects are detected and documented by experts on the field using paper-based surveys. However, these methods are time-consuming and require immense expertise and cost. This paper presents the novel approach of condition survey by automated defect detection system to overcome these limitations. The automated defect detection system is developed using the Faster R-CNN (Region-based Convolutional Neural Network) to detect multiple types of defects from the heritage monument photos. Moreover, the web-based interface is developed to run the automated condition survey system. Types and number of defects and the position of defects are the output of the developed condition survey system. The system is demonstrated for the national heritage monument located in Surat city to check the developed system's effectiveness and field application. The proposed system is automatic, fast, and reliable for a condition survey of heritage monuments. This system will be useful for conservation, preservation, and maintenance management of heritage monuments.KeywordsCondition surveyAutomatic defect detectionHeritage monumentsDeep learningFaster R-CNN