<p>Traditional Direction-of-arrival (DOA) estimation methods based on Fractional Fourier transform (FRFT) are computationally demanding and ineffective at low signal-to-noise ratios (SNR) or insufficient number of snapshots. To address this issue, the paper proposes an improved DOA estimation method based on FRFT. The proposed method utilizes the aggregation characteristics of FRFT for Linear Frequency Modulation (LFM) signal and derives the linear relationship between the peak position on the optimal Fractional Fourier (FRF) domain and the position of different array elements. Subsequently, by transforming the two-dimensional DOA estimation problem into a linear fitting of slope problem, the slope is fitted by the batch gradient descent method, and the two-dimensional arrival angles are obtained using the derived relation. The validity of the estimation is demonstrated by further analyzing the Cramér-Rao Bounds (CRB) for arrival angles in L-shaped array. The effectiveness of the proposed algorithm is verified through comparing with other algorithms, and it is also shown that the algorithm has high estimation accuracy even in low SNR or less number of snapshots.</p>