The work is dedicated to the task of autonomous visual navigation of a Mars rover. The current stage of space exploration is closely linked to the study of nearby planets in the solar system using robots with varying degrees of autonomy. The distance of these planets from Earth causes significant signal propagation delays, complicating the use of teleoperation and making the use of autonomous vehicles promising. A review of existing approaches and methods for three-dimensional reconstruction of the observed surface in the task of autonomous navigation of a prospective Mars rover has been performed. A classification of approaches and methods is provided, their properties are considered, their applicability to the task at hand is evaluated, and possible options for building a visual navigation system are described. The work serves as a foundation for the further development of the visual navigation system of the Mars rover.