Skip to Main Content
This paper presents a new model-free visual servoing that is able to servo a robotized 2D ultrasound probe that interacts with a soft tissue object. It makes direct use of the B-mode ultrasound images in order to reach a desired one. This approach does not require the 3D model of the object nor its location in the 3D space. The visual features are based on image moments. The exact analytical form of the interaction matrix relating the image moments variation to the probe velocity is modelled. To perform model-free servoing, the approach combines the image points coordinates with the probe pose to estimate efficiently 3D parameters required in the control law. The approach is validated with simulation and experimental results showing its robustness to different errors and perturbations.