Pomelo is a fruit with high annual output and wide popularity in China. Currently, the pomelo processing industry is in a stage of rapid development, and the processing potential of pomelos is gradually being exploited. However, the step of adjusting pomelos to a uniform orientation in the processing flow still relies on costly and inefficient manual labor. To achieve fully automatic orientation adjustment of pomelos, reduce costs, increase efficiency, and liberate human labor, this paper proposes and designs an automatic orientation adjustment device for pomelos based on machine vision: images are collected through the cooperation of cameras and infrared sensors; the OpenCV library is used to write image processing programs to determine whether the pomelo needs to be adjusted; the shape is simplified and the adjustment parameters are calculated; the adjustment rod is controlled to contact the pomelo, and the pomelo completes the orientation adjustment through planar motion. Eventually, the pomelos leave the device with a uniform orientation. Through experiments on the common Shatian pomelo in South China, the results show that under the two working modes, the accuracy rate of initial posture recognition and the average success rate of orientation adjustment for Shatian pomelos are 93.3%, 82.6% and 96.7%, 85.1% respectively. The average operation time of the program for pomelo orientation adjustment in the parallel working mode is 5.14 seconds. The proposed pomelo orientation adjustment method not only helps to improve the working efficiency of the Shatian pomelo processing industry, but also has the potential to be applied to the orientation adjustment of other varieties of pomelos.