The paper proposes a new approach that enables the structure analysis and reconstruction of a rough surface where the height of inhomogeneities (from the depression to the upper point) varies within the spread about 20 nm. For the surface diagnostics, carbon nanoparticles are used, which serve as sensitive probes of the local surface height. A single nanoparticle can be positioned at a desirable point of the studied surface with the help of an optical tweezer employing the He-Ne laser radiation. Then the particle is illuminated by the strongly focused exciting beam of 405 nm wavelength, with the waist plane precisely fixed at a certain distance from the surface base plane. The particle’s luminescence response (in the yellow-green spectral range) strongly depends on the distance between the exciting beam waist and the particle, thus indicating the local height of the surface. After scanning the surface area and the consecutive interpolation, the surface “vertical” landscape can be reconstructed with a high accuracy: the numerical simulation shows that the RMS surface roughness is restored with an accuracy of 6.9% while the landscape itself is reconstructed with the mean error 7.7%.