INTELLIGENT DETECTION OF CONCRETE APPARENT DEFECTS BASED ON A DEEP LEARNING - LITERATURE REVIEW, KNOWLEDGE GAPS AND FUTURE DEVELOPMENTS

Abstract

A variety of apparent defects in the development process and use of concrete in structures weakens the integrity and long-term performance of concrete structures; therefore, apparent defects in concrete need to be efficiently and intelligently detected. Compared with the traditional manual detection, the intelligent detection of apparent defects based on deep learning can significantly improve the efficiency and accuracy of the detection, and provide a new technical means for the detection of concrete structure defects. At the level of technical applications, this paper summarises the application status of deep learning algorithms in the detection of apparent defects in concrete from the three aspects of image classification, target detection and semantic segmentation and highlights the shortcomings of the current applications. This paper introduces the defect image acquisition method from the two aspects of manual photography acquisition and UAV acquisition and summarises the construction and expansion methods of the dataset from the two aspects of the annotation method and data enhancement technology. The problems and directions for the wide application of deep learning in intelligent identification of apparent defects in concrete are analysed and discussed.

Keywords

• concrete defect
• artificial intelligence
• image classification
• target detection
• semantic segmentation