In this paper, we present a work that deals with the performance of robotic tasks through 2D photometric visual servoing methods. These approaches skip the usage of the extracted geometric features and instead use visual features based on pure luminance intensities. Unfortunately, these methods are affected by the change of vision especially illumination changes and partial occlusion. Any change in the conditions of the newly captured images directly alters the servoing task. To overcome these problems, we use Gabor features, which are known for their invariance and robustness and can thus resist local distortions caused by illumination variations, deformations, and so forth. Gabor features optimally represent the time-frequency content the expanded input image using a bank of four Gabor filters with different angles and frequencies as expansion functions, and are used as signal control inputs. The interaction matrix development for the proposed features is presented. Simulation results prove the robustness and the effectiveness of this technique against occlusion and illumination variations.

Visual servoing, Gabor filters, Illumination variations, Interaction matrix.