Abstract: A second-order nonlinear filter which can estimate the attitude of a three-axis stabilized gyroless satellite from star sensor vector observations is developed using the nonlinear filter technigue based on second-order Taylor series approximation. Considering the characteristics of vector observation, the methods to dispose vector observation information in the case of one vector and multi-vector are discussed separately. In addition, the QUEST algorithm is efficiently combined into the attitude estimator as the compressing technique of multi-vector observation information in order to simplify the update implementation. Performance comparison made by computer simulation demonstrates that the second-order nonlinear attitude estimator consistently performs better than the traditional extended Kalman attitude estimator.