A high precision crack classification system using multi-layered image processing and deep belief learning. Issue 2 (1st February 2020)

• Record Type: Journal Article
• Authors: Jo, Jun; Jadidi, Zahra
• Is Part Of: Structure and infrastructure engineering. Volume 16:Issue 2(2020)
• Page Start: 297
• Page End: 305
• Publication Date: 2020-02-01
• Subjects: crack detection -- image processing -- Hessian filter -- Gabor filter -- Otsu filter -- Retinex filter -- Median filter -- deep belief network -- classification; Structural analysis (Engineering) -- Periodicals; Structural engineering -- Periodicals; Buildings -- Performance -- Periodicals; 620.005
• DOI: 10.1080/15732479.2019.1655068
• Languages: English
• ISSNs: 1573-2479

Abstract

Abstract: Road surfaces experience fatigue stress and loading, which often lead to cracks on the surface. The cracks might cause serious damage, and therefore, early detection can reduce the road maintenance cost. Traditional inspection methods are carried out by humans and are slow, costly and hazardous. To improve accuracy and reduce the hazards of current crack detection methods, this paper proposes a new autonomous crack detection system (ACDS) that can be used in any autonomous vehicles (UAVs). ACDS consists of three stages: image acquisition, image processing, and classification. The image processing stage consists of five parallel filtering methods, which remove noise and extract features from the images. In the classification stage, five deep belief network (DBN) classifiers separately analyse the images to detect cracks. The dataset used in this paper contains 15, 000 RGB and infrared images, with or without cracks. The results show the high precision of the proposed system.