Abstract
In the past years there has been an extensive effort to develop an automated crack detection method by image processing to improve inspection and evaluation process of concrete structures. However, these methods are not yet accurate enough due to the difficulty and complexity of the problem. Especially, the mold mark is misjudged as the crack because image characteristics are quite similar to each other. To solve this problem, this paper proposes the method which distinguishes cracks and mold marks properly by convolutional neural network which is a type of deep learning. Then, accurate classification is archieved by the random forest method with considering image characteristic related to pixel value and geometric shape. The accuracy of developed method is investigated by the photos of concrete structures with lots of adverse conditions including not only the mold mark but also shadow and dirt, and it is found that the method can extract the crack region with high accuracy.
