A running vehicle equipped with detection equipment can efficiently monitor road conditions over a wide area. The generated massive amounts of detection data require timely processing to provide essential information for road safety monitoring and maintenance. Vehicular edge computing is able to offload detection data processing tasks to nearby intelligent vehicles or roadside units, alleviating the processing pressure on detection vehicles with insufficient computing power. However, the uncertain movement of detection vehicles due to traffic light control, road congestion, etc. may lead to task offloading transmission interruption and even unpredictable changes in edge computing service relations, thus resulting in a waste of edge network resources. To address this problem, this paper proposes a risk aversion driven computation offloading scheme for vehicle-mounted detection equipment with uncertain mobility behavior. We introduce a risk assessment based offloading utility model and leverage non-dominated sorting genetic algorithm II to design a task offloading and resource joint scheduling scheme. Simulation results demonstrate that our scheme outperforms benchmark schemes in task offloading efficiency and data processing delay.