Community policing to collaboratively maintain community safety and order in conjunction with law enforcement is becoming increasingly popular and efficient with the use of mobile technologies. Beyond sharing information about local problems such as crime via online discussion forums, there has been an increased focus on the impact of mobile crowdsourcing systems on community policing. In this study, we designed a novel crowdsourced patrolling campaign in which community members schedule their own patrol times and routes, then perform bike-based patrolling with video capturing using their smartphones. We conducted a four-week field study (n=20) on a university campus to verify the campaign's feasibility and observe users' behavior. Our results show key findings about users' task scheduling, event capturing and reporting behaviors, factors affecting task selection and execution and user motivation and engagement. Finally, we discuss several practical design implications in building crowdsourcing systems for community policing.