An automatic alignment device for pomelos based on machine vision was designed to achieve fully automatic alignment of pomelos to reduce costs， increase efficiency， and set free manpower. A camera and infrared sensors were used in conjunction to capture images. OpenCV was used to write an image processing program for determining the alignment of pomelos and simplifying shape calculation of alignment parameters to complete the task of controlling the planar motion of pomelos and achieve a unified alignment to leave the device through the aligning rod. The designed device was tested on Citrus maxima （Shatian pomelo） in Guangdong region. The results showed that the device achieved the accuracy rate of initial posture recognition and the average rate of successful alignment for Shaddock pomelo in the serial and parallel operation mode was 93.3% and 82.6%， and 97.4% and 87.5%， respectively. The average operation time for the program to align pomelos in the parallel operation mode was 5.20 s. It is indicated that the automatic alignment device designed for pomelos can accurately identify the pomelos that need to be aligned， and efficiently carry out the task of aligning pomelos. It will provide a feasible solution for automatic alignment of pomelos.