Because the surface micro-topography has a large impact on soil properties, numerous studies have focused on surface roughness, soil height changes, and soil cloddiness characterization. Usually, feature parameters are estimated from soil measurement samples, based on statistics characterizing the surface as a whole. Now, the shortcoming of such a global approach is that it fails to detect local soil height changes and non-stationarities. The present study introduces a new method to identify and characterize the clods on a seedbed surface Digital Elevation Model (DEM). It is based on an a priori model of the clods, namely objects presenting closed elevation contours with high gradient values. Our clod segmentation method was assessed with the help of a soil scientist on the two kinds of tilled soil surfaces which were considered in this study: an artificial surface made in the laboratory to have a controlled roughness, and an actual seedbed surface made in an experimental field. In both cases results were evaluated in terms of sensitivity and specificity, and showed the performance of the method. We also study the impacts of the main parameters of the method and the computer time. Its main limitations are that it fails to identify the small clods (diameter smaller than 7 mm in this study) and the clods embedded within another piece of relief, such as a greater clod or a hollow border. Then, we show how results can be used to compute clod parameters: mapping of clods provided as output of our algorithm and clod shape measurements. Finally, an application to study soil heights changes with rainfall is provided.