Check dams are widely constructed on the Loess Plateau as effective engineering measures to prevent water and soil loss. Accurately extracting the location of the check dam and estimating the remaining storage capacity are crucial for assessing the operational status of existing dams and planning the spatial distribution and scale of new dam construction. Digital elevation model (DEM) supports data analyses by providing a digital representation of surface morphology. Therefore, this study proposed a method (feature line detection combined with continuous gully network, FLD-CGN) to extract dam information. The FLD-CGN method first traverses each D8 drainage network cell as the initial growth cell, employing a two-level nested breadth-first search strategy combined with specific rules to expand the region and extract wide gully watercourses, which are then processed morphologically to obtain the gully network. Second, dam edge line detection algorithms are designed to extract all dam edge lines with complete and independent features. Finally, FLD-CGN uses the multiresolution segmentation algorithm to extract the potential sedimentation areas of the dam. It builds a dam framework based on the spatial relationship between the edge lines and the gully network to determine the dam location. The Wangmaogou and Zhoutungou watersheds in Shaanxi Province, characterized by loose soils and severe water-soil loss, were selected as representative study areas. In these regions, the FLD-CGN method achieved correctness and completeness rates of 92.5% and 90.2%, respectively, and further estimated the remaining storage capacity. Among the 23 check dams in Wangmaogou, 16 still possess silt retention capacity, with a total remaining storage of 127.1008 x 104 m3; in Zhoutungou, 15 dams retained 56.2421 x 104 m3. This study supports the scientific analysis of the spatial layout planning of check dams and the management of water and soil loss on the Loess Plateau.