Abstract: A merit function based on modified spatial frequency domain(UV) sampling method is proposed, which can be used to optimize the geometny of long baseline astronomy optical interferometer telescope array. The UV region to be sampled is firstly divided in both radial and rotational directions, then the number and density of UV sampling points of each segment are calculated. The deviation of UV sampling point density distribution from the ideal Gaussian distribution is taken as the merit function value. The genetic algorithm is used as the optimization algorithm, due to its global convergence capability which can reduce the dependence on initial array. A 6 apertures telescope array was optimized with the proposed merit function using genetic algorithm. Moreover, the optimized array(Array-6 array) was compared with the CHARA array. Analysis results show that the Array-6 array UV sampling points distribution in the radial direction is intensive in low frequencies and continuous, which is beneficial to the reconstruction of observation target contour and structure. The binary star observation and image reconstruction simulations show that the array optimized using the proposed merit function has better imaging performance, compared with the CHARA array. The error of reconstructed image using Array-6 is 21.34, which is 18.16% smaller than the error of reconstructed image using CHARA. Additional, the proposed merit function can be used to optimize arrays with large aperture numbers, so it also has a large application prospect for the optimal design of the radio-band telescope array.