The receiver array of matched field processing usually needs to cover the whole water column, such a large arrangement of array hinders its application seriously. To overcome this shortcoming, an approach called matched field processing based on least squares algorithm is proposed, which decomposes the received fields into depth function matrixes and amplitudes of normal mode at the beginning. Then all the mode amplitudes are estimated using the least squares in the sense of minimum norm. Next, the amplitudes estimated are used to reconstruct the received fields of the short vertical linear array, which makes the reconstructed ones contain more environmental information, therefore the performance of matched field processing with short array is improved. In the end, lots of simulations with three short vertical arrays which are located near by water surface are processed in the classical Pekeris waveguide, and the performance of matched field passive localization is evaluated. Meanwhile, in order to quantify the influence of different kinds of geometrical arrangements to the received fields, the generalized cosine-squared between two vectors is used. The results show that when the received fields of short array are reconstructed using the least squares, the generalized cosine-squared is larger obviously, finally the performance of matched field passive localization with short array is improved and the proposed algorithm is proved to be effective.