Poster: Crowdsourcing for video traffic surveillance

Video traffic surveillance monitors traffic situations such as traffic jams, traffic accidents, or running a red light. Although automatic traffic event detection has been studied for years, current systems often fail to handle various situations and do not fully take advantage of existing video traffic surveillance data. Hence, there is a need for an approach that integrates labor resources with intelligent video analysis to enhance the robustness of video analysis models and fulfill the demands of traffic surveillance. Motivated by the intuition that a driver or pedestrian often needs to know the exact traffic conditions before selecting a particular route, we propose a crowdsourcing [2] surveillance framework to assist existing traffic surveillance systems. In particular, people can use their smartphones to check the detected traffic situation and the corresponding video clips received from the video surveillance system, and make quick judgements about the received results. This finegrained information provided by traffic surveillance system not only shows the detected traffic results but also presents live video clips. Furthermore, smartphone users can provide their feedback to the system for improving the intelligent video surveillance model or correcting errors that may be present in the current traffic event detection.