Detecting cracks on a concrete surface using histogram of oriented gradients

Abstract

With the aging of structures such as bridges, buildings, and tunnels, investigating the cracking of concrete surfaces has become very important. The standard method for investigating cracks has usually been to perform a visual inspection and acoustic diagnosis. However, the inspection results when using these techniques vary depending on the experience of the inspector, causing there are not objectivity and quantifiability enough. Therefore, digital image analysis of concrete cracks is being developed with the aim of providing efficient, automatic, and quantitative inspection. However, in the process of binarization, some of the cracks become fractured or invisible, which has a negative effect on the investigation accuracy. To improve the accuracy when detecting cracks, we propose a method that uses the histogram of oriented gradients (HOG) to detect cracks on the concrete surface. We are also working on the use of unmanned helicopters to take photos of the concrete surfaces to enable automatic investigation.