Pipelines are important infrastructures to support our normal lives. In modern cities, most of them are laid underground along the roads. Some pipelines carry hazard contents and therefore are very dangerous. However, accidents did occasionally happen and resulte in severe consequences. In order to protect the pipeline infrastructures and surrounding people, preventing disaster from happening is indispensable. Recent reports showed that most pipeline failures were caused by the third-party activities, mainly constructions, rather than the material failure or corrosion. Environmental surveillance is therefore proposed to detect the constructions around the pipeline and provide early warning of the danger. In this paper, constructions are detected based on the detection of the construction machines. The recognition of the most representative construction machine, road cutter, is detailed studied. The windowed average power spectra density (WAPSD) is proposed for the feature vectors. One-class support vector machines (SVM) will be used for the classifier. Extensive on-site experiments were conducted to verify the feasibility and effectiveness of the proposed method. The analysis results show that construction activities around pipelines can be detected effectively. Thus the possible disaster can be avoided and the integrity of the pipeline infrastructure can be protected.
展开▼
