Optimized Deep Encoder-Decoder Methods for Crack Segmentation

Continuous maintenance of concrete infrastructure is an important task which is needed to continue safe operations of these structures. One kind of defect that occurs on surfaces in these structures are cracks. Automatic detection of those cracks poses a challenging computer vision task as background, shape, colour and size of cracks vary. In this work we propose optimized deep encoder-decoder methods consisting of a combination of techniques which yield an increase in crack segmentation performance. Specifically, we propose a new design for the decoder-part in encoder-decoder based deep learning architectures for semantic segmentation. We study its composition and how to achieve increased performance by exploring components such as deep supervision and upsampling strategies. Then we examine the optimal encoder to go in conjunction with this decoder and determine that pretrained encoders lead to an increase in performance. We propose a data augmentation strategy to increase the amount of available training data and carry out the performance evaluation of the designed architecture on four publicly available crack segmentation datasets. Additionally, we introduce two techniques into the field of surface crack segmentation, previously not used there: Generating results using test-time-augmentation and performing a statistical result analysis over multiple training runs. The former approach generally yields increased performance results, whereas the latter allows for more reproducible and better representability of a methods results. Using those aforementioned strategies with our proposed encoder-decoder architecture we are able to achieve new state of the art results in all datasets.