Video streaming from cameras, closed-circuit television (CCTV), smartphones and Internet-connected objects (ICO) largely contributes to big data traffic on the Internet. Video streaming provides enormous amount of useful information for delivery of efficient and effective services in smart cities. Modern cities have large networks of surveillance cameras including CCTV, street crossings and the like. In this paper we discuss the challenges of annotating and retrieving video data streams from vehicle-mounted surveillance cameras.We also propose and evaluate the CityWatcher application – an Android application for recording video streams, annotating them with location, timestamp and additional context in order to make them discoverable and available to authorized Internet of Things applications. One of such applications is based on crowdsourced alerts to city authorities about road problems, like potholes, cracks, traffic accidents. These alerts are driver-initiated and are rewarded through an incentive mechanism. OpenIoT platform is used for infrastructure and development support.