Recently, there has been attention to smart city and the increasing of labour costs, how to continuously and automatically monitor the structure with the least amount of manpower has become an important research direction. Therefore, there are tremendous robotic-based applications designed to observe, predict, or detect the changes that could happen in the civil engineering infrastructure. One of these applications is detecting road cracks. In this project, an overview is conducted to explain the automatic crack detection systems then we present a detailed analysis about extracting the road cracks. In particular, U-Net is one type of Fully Connected Convolutional Neural Network that can be applied on various of road crack datasets. U-Net is faster because it needs a small number of images for training. Also, it can detect cracks under various conditions such as background noise, shadow, illumination, etc. Traditional image processing methods are combined with U-net model to improve the performance through focus on region of interest. This research can be benefited to control the traffic speed, grantee safety on roads and regular maintain civil infrastructure.