Concrete spalling and crack inspection is a labor intensive and rout ne tasks, which is more challenge with bridges hard to access. Whereas the tremendous recent pro- gresses in concrete inspection are based on even surface under ideal illumination, and no open database was released so far. In this paper, we propose an online inspection and image collection system using Unmanned Aerial Vehicle (UAV) (Fig.1) besides the web-exploration approach. We introduce a new Concrete Structure Spalling and Crack database (CSSC) using web-exploration over 38,483 images. Then, the UAV deploys the trained model for inspection, and collects images which contains region of interests (ROI) with possible flaws. Thus extra 7,648 images were collected through this approach to assist further training. We illustrate the complete procedures to do labeling, training, and post processing to find the corresponding ROI with VGG-16 . We also provide a comparison on the database generated before and after field collection. Experiments on field data show that the proposed approach provides a robust visual inspection solution for concrete bridges and it is venerable for light illumination.