Range estimation using TDL neural networks and application to image-based visual servoing

In this paper, a new approach is proposed to estimate the depth of the target using active monocular stereo. The proposed method employs a tapped delay line (TDL) neural network to approximate the depth of the target. It is shown that the proposed method is less computationally expensive and functions well in case of occluded or unmatched features making it more robust than similar methods such as homography-based techniques. Experimental results validate the robustness and accuracy of the approach. The application of the proposed method to image-based visual servo (IBVS) of a 5 degrees of freedom manipulator is discussed.