Pavement crack detection based on a deep learning approach and visualisation by using GIS

ABSTRACT

One of the most prevalent ailments of the road is crack. When it does, the standard of road engineering will be significantly lower and may even result in road collapse. Early detection of the cracks will result in significant maintenance cost savings if timely maintenance is performed. Although direct detection is highly challenging, the range of image cracks on the actual pavement is too large, the image quality is insufficient, the composition is complicated and the image range is too wide. Traditional manual detection methods have a number of drawbacks, including a lack of precision, a significant danger of detecting operation and a lengthy processing time. Therefore, in this research, we employed a novel technique based on image processing for the pavement detection method based on the features of pavement cracks. In order to avoid those difficulties Pre-processing (Median Filter), feature extraction (ResNet-50) and feature selection (Adaptive Salp Swarm Algorithm) techniques are also included in order to produce a clear image. With this proposed method, a variety of crack types including longitudinal, transverse, alligator and block can be identified.

KEYWORDS:

• Pavement crack detection
• median filter
• ResNet50
• adaptive salp swarm algorithm
• deep belief network (DBN)
• rider optimisation algorithm (RO)
• GIS (geographical information system)

Disclosure statement

No potential conflict of interest was reported by the author(s).