In order to get a more precise 3-D reconstruction, a peak detection method based on Gaussian fitting of trust region has been proposed in this work. Gaussian modeling is performed on the returned wave of single time channel of each frame, then the modeling result which can effectively reduce the noise interference and possesses a unique peak could be taken as the new returned waveform, lastly extracting its feature data by peak detection. The experimental data of aerial target is for verifying this method. This work concludes that the peak detection method based on Gaussian fitting reduces the extraction error of the feature data to less than 10%; ultilizing this method to extract the feature data and reconstruct the target makes it possible to realize the spatial resolution with a minimum 30 cm in the depth direction, and improve the 3-D imaging accuracy of the Streak tube imaging lidar concurrently.