A work machine is provided with grade control capability using an imaging system, e.g., rather than GPS. The work machine includes at least one work implement for working at least part of a terrain, and first sensors (e.g., cylinder sensors) generate signals corresponding to positions of the work implement. Second sensors (e.g., stereo cameras) generate signals corresponding to positions of representative features of the terrain (e.g., curbs) in a field of view. A controller receives the signals and determines in a local reference system independent of a global reference system: first position information corresponding to the work implement; and second position information corresponding to the representative features. According to a selected control mode, target parameters for the work implement are determined based on the second position information corresponding to the representative features, and output signals are generated corresponding to a difference between the first position information and the target parameters.