Bridge damage detection is of vital importance to bridge safety. Nowadays the damage detection is mainly performed by human which is inefficient. We pro-posed a bridge damage detection and recognition method based on deep learning which is named DT-YOLOv3 in this paper. Our method is based on YOLOv3 object detection method and several improvements were made. First, deformable convolution was used to extract more accurate features, and transfer learning was introduced to improve the detection accuracy. Then, the model was compressed using group convolution and pruning. The test results show that our method is more effective than state-of-the-art methods and costs less time.
Present video retrieval methods have many problems. To solve these problems, a new video retrieval algorithm base on the combination of video spatio-temporal feature curves and key frames is proposed in this paper. In this new algorithm, the feature curves are extracted from the video, and then two videos' feature curves are compared to determine whether they have the same content or not. In the comparing process, to solve the problems of brightness offset in all frames and abrupt intense disturbance, the paper uses Grads comparison method and Exception Factor. To improve retrieval precision for videos like films, a method based on clips and key frames is added after spatio-temporal feature curves based retrieval. The new algorithm can largely reduce the data amount of video retrieval. Experimental results show the effectiveness and robustness of the algorithm.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.