A sensor management method for the combined task of target detecting and target tracking is studied based on risk theory. Firstly, the target detecting risk and target tracking risk are separately defined, and the computing methods are given. Then, minimizing the total of the target detecting risk and the target tracking risk is seen as the objective function, to build the sensor scheduling model. What is worthy of emphasizing is that the sensor scheduling takes place on the sensor network which is deployed by the preceding sensor deployment model. To get solutions for the two models, and to overcome the artificial bee colony algorithm's defect of easily striking in local optimum, the improved bee colony algorithm based on two-way roulette and particle swarm is proposed. Finally, simulations are carried out to illustrate effectiveness of the proposed models and algorithm.