To solve the problem that vector reconstruction method with uniform linear arrays cannot be used directly in uniform circular arrays, an efficient vector reconstruction algorithm based on space mode for DOA estimation is proposed. The eigenvector corresponding to the largest generalized eigenvalue of the covariance matrix is corrected to acquire signal eigenvectors after mode excitation. The receiving data matrix is constructed by the forward-backward vector reconstruction to estimate DOA of coherent signals. Optimization algorithm is presented to acquire the largest signal subspace eigenvector by performing eigen-decomposition before mode excitation, which eliminates the noise fully and avoids the generalized eigen-decomposition on the virtual linear arrays, and the computation complexity is reduced obviously. The theoretical analysis and numerical examples are provided to demonstrate effectiveness of the proposed approach.