In this article, a theoretical method based on the fluctuation of gradient tilt (G-tilt) of active light source is proposed to estimate the horizontal profiles of atmospheric optical turbulence (Cn2) and transverse wind. The G-tilt, related to the average phase gradient, is in the same direction as the average ray direction. And G-tilt angle is considered to be equal to the ratio between the centroid position offset and the focal length. In this method, a theoretical model based on lidar system is set up, in which forward scatter light beams at different distances are taken as beacons. These beacons are detected by a two-aperture telescope. And two light columns, from which we can obtain the information about G-tilt angle, are imaged by these beacons. In order to obtain the turbulence intensity and wind velocity from G-tilt angle with our theoretical model, the differential cross-correlation expressions of G-tilt angle and its derivative are derived in detail. These two expressions are based on the spatial cross-correlation function obtained from Rytov approximation and Taylor's frozen-flow hypothesis for Kolmogorov turbulence. Simultaneously, path weighting functions of Cn2 and wind velocity are derived, and the effects of path weighting functions on the calculation of our method are analyzed. Based on such an analysis, to realize the inversion of turbulence intensity and transverse wind, the matrix transformation algorithm is proposed. We ignore some minimal values of the path weighting functions in our algorithm so that the ill-conditioned matrix is avoided. Besides, numerical simulation is used for preliminarily validating this method. In our simulation, Cn2 varies randomly between 10-15 m-2/3 and 10-14 m-2/3 while wind velocity ranges from -5 m/s to 10 m/s. The sign of the wind velocity represents the direction of wind. According to the simulation results, the horizontal profiles of atmospheric optical turbulence and transverse wind calculated are consistent with their theoretical values no matter whether there exists Gaussian noise. When the ratio between the standard deviation of Gaussian noise we added and the original signal is 0.2, the maximum relative error of logarithmic Cn2 is 3.4%, and the correlation coefficient between the calculated results and theoretical values for Cn2 is 0.8. Besides, the maximum absolute error of wind velocity is 1.82 m/s, and the correlation coefficient between the calculated results and theoretical values for wind velocity is 0.9. Even if the horizontal profiles of atmospheric optical turbulence and transverse wind vary largely, the calculation results of our method remain stable. Therefore, a new idea is provided for measuring atmospheric turbulence and wind.