Photometric Visual Servoing Through Sobel-Based Image Gradient Utilization

This paper deals with the development of a new 6-degree-of-freedom (DOF) visual servoing control law. The traditional visual servoing (VS) methods depend on geometric features to design the control law, which limits their versatility due to reliance on visual tracking algorithms. To address these limitations, direct visual servoing (DVS) approaches have been introduced, showing that the design of photometric visual servoing (PVS) can bypass geometric feature extraction by directly considering the luminance of all image pixels to control the robot. However, these methods are sensitive to illumination changes and partial occlusions, which make the control task non-robust. To overcome this, we develop a new direct visual servoing control approach based on a Sobel filter to enhance the precision of image information under changing lighting conditions by extracting image gradients. These gradients are then used to design the interaction matrix that allows real-time updates to the motion of the robot for adaptability and precision. Also, the proposed control scheme has been tested on the VISP platform and was compared to the classical photometric visual servoing in order to evaluate its efficiency in nominal and unfavorable conditions. Experimental results validate that the approach provides more performance and reliability under variable illumination conditions.