In the past several years, UAV has been extensively used in agriculture. However, the efficiency is still not as high as desired and the phenomenon of pesticide pollution is still existing. This is mainly because of the following two problems: 1) the autonomy of most existing UAV system is still very limited. Actually, most of them are still operated through remote-controlling. 2) the UAV's operating precision is not high enough due to the low accuracy flight control near the plants. Therefore, in 2016, the Aviation Industry Corporation of China (AVIC) has announced a UAV competition, called International Unmanned Aerial Vehicle (UAV) Innovation Grand Prix (UAVGP), to promote the technique progress of agriculture oriented UAVs. This competition mainly focuses on a typical high precise spraying mission, and in order to present the autonomy, all involved UAVs are required to operate autonomously without any human intervention. Our UAV system finally won the championship of this competition. In this paper, we present the whole system design firstly; after that, the core algorithm of task-scheduling and vision for identification and localization are given; Finally, the results of experiments and competition are introduced and analyzed in detail to show the efficiency and precision of the whole system.