Precise path control of unmanned aerial vehicle (UAV) is the key technology to realize aviation plant protection in mountain orchards. In order to carry out the plant protection operation stably and accurately in mountain orchards, based on the four-rotor UAV, the flight path control system and method of UAV for plant protection by fusing GNSS and machine vision were designed. The system consisted of two parts, i.e., UAV flight platform and ground control station. The former consisted of a four-rotor UAV, an inner-ring flight control module, a GNSS moving station, a RGB camera, a wireless video transmission module and an electronic compass. The latter consisted of a GNSS base station, a flight control module, a laptop and a video capture module. A fruit tree row recognition algorithm was proposed based on Python language combined with OpenCV library. In this algorithm, a linear combinatorial algorithm was used to extract the target area, and the least squares method was used to fit the central point of target area to obtain the row trend of fruit tree. Then the yaw angle was calculated to realize the path control. In the working process, the UAV was controlled by the visual navigation method when flying over the fruit tree, while controlled by the GNSS navigation method when it needed to switch between rows of different operation fruit trees. The developed system and proposed method was tested in mountain apple orchard. The results showed that when the flight speed was 2m/s, the height of UAV from fruit tree canopy was about 2m, the camera’s dip angle was 46° and the image navigation control rate was 2 times/s, the absolute path control error of the system was -47~42cm, and the average absolute error was -9cm. The high control precision indicated that the system could meet the requirements of UAV for plant protection operation in mountain orchards. The research provided a new method for path control of UAV on plant protection in mountain orchards.