The principle of differential column image motion light detection and ranging for acquiring atmospheric turbulence profile is described. Aiming at the large retrieval error in high-altitude turbulence of current Levenberg- Marquardt inversion algorithm, a novel inversion model with inequality path constrained is developed and the penalty function method is used to handle this model, thereby an unphysical solution by adding the information of highaltitude turbulence is avoided. Furthermore, in order to weaken the current algorithm dependence on initial value and priori knowledge, a new optimization strategy based on genetic algorithm is presented to locate initial value of current algorithm in global variable space. Typical atmosphere turbulence profiles are simulated with the modified algorithm and the current algorithm. The measured lidar data in Hefei is also analyzed. Results show that the modified algorithm can enhance the global search capability of iteration process and perform strong robustness against measurement noise, improving the retrieval precision and accurate quantification of high-altitude turbulence effectively. Moreover, the modified algorithm accelerates the convergence.