An optimization model of circle array was set up from the basic optical synthetic aperture imaging principle. The circle array was optimized by adopting a genetic algorithm with an improved real coding method coding the location of sub-apertures. The measure function was designed based on maximizing the distances between u-v coverage dots and minimizing the redundant array. The point spread function, optical transfer function and diffractive imaging were analyzed with the circle array synthetic aperture imaging system. The optimized result of 8 to 16 sub-apertures on a circle array was obtained, and they were compared to the results achieved through simulated annealing algorithm. Using the emulator program, the point spread function was analyzed and contrasted to that of a uniform circle array. Results show that the real coding genetic algorithm can resolve the array optimization well, cost less time and get a better optimization compared with the simulated annealing algorithm.