Practical methods of image data preprocessing for enhancing the performance of deep learning based road crack detection

초록

This paper proposes practical preprocessing methods for effective training of the deep learning model, taking into consideration the characteristics of image data used to detect cracks on the road surface. Convolution deep neural network extracts and recognizes features of a region included in an image, and classifies the image based on the features. However, collected raw image data do not guarantee the effective training of the deep learning model because the characteristics of the crack region and the non-crack region are not clearly distinguished and irregular noises are included in the raw image data. Therefore, we propose the preprocessing methods that can improve the performance of deep learning model in order to solve these problems. © ICIC International 2020.

키워드