The Gobi desert is one of the most widespread landscapes in northwest China. The coverage and morphological characteristics of gravels on the Gobi desert are often used to understand the evolution process of Gobi desert. However, some traditional methods e.g. eye estimating and sieving, can’t accurately measure these indices and is time-consuming. Here we develop a new method to accurately, promptly and automatically analyze the gravels characteristics of Gobi desert from orthoimages using decision tree model, watershed segmentation and hierarchical analysis. The main results list here including: 1) The nearer to the source region of deposits, the larger gravel coverage is. Gravel coverage increased 16% as elevation increased 1000 m along the alluvial fan. 2) Fine gravel content increased with the decrease of elevation. The average of gravel diameter is 13 mm at 820 m of elevation and 17 mm at 1840 m of elevation. Comparing the gravel morphological characteristics calculated by proposed method with manual digitization of images, the calculation time of the proposed method is 60 times faster and the calculated indices have a good agreement. Therefore, the proposed method provides a fast and precise approach to quantify the gravel morphological characteristics of Gobi desert and explore the source region of deposits and the transport/deposition process in the formation and evolution of Gobi desert.