Forest stand segmentation is a critical process for forest management and inventory. The forest stand segmentation accuracy will determine the forest stand level parameters quality. In this study, we developed an automatic forest stand segmentation algorithm based on ArboLiDAR, a software used to process Light Detection and Ranging (LiDAR) point cloud data. We then optimized the parameters for the algorithm to the Dayekou forest area on Qilian Mountain in China to find the most suitable parameters for automatic stand segmentation. Further, we extracting the forest parameters at the stand level based on Bysh method. Our results showed that the limited region growing method based on the gradient is the most suitable one for analyzing automatic stand segmentation in the studied area. Among our tested parameters groups, the fifth group contains the optimal parameters for the studied area. In addition, for forest parameters, the R2 of mean height (H), average diameter at breast height (D), basal area (G), and Stand volume (V) is 0.744, 0.720, 0.562, 0.696, respectively. The RMSE value is 5.24%, 28.57%, 19.93%, and 17.66%, respectively. Our study serves as a technical basis and reference for future studies that perform more efficient analyses on forest resource inventory in China.