Depth estimation from monocular vision using image edge complexity

Abstract

Autonomous robotic arm motion requires the use of a control system in order to prevent collisions with the targeted object. Generally, in translational motion, as the camera approaches an object, the degree of complexity of the edges of the object image will change. This principle can be used to estimate the distance to a targeted object. This work introduces a novel statistical method, named Moment of Zoomed-Algorithm Kurtosis (MoZAK), which is based on the I-kaz method, as an indicator for motion system control. The MoZAK parameter, ℒ <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">c</sub> which represents the degree of complexity of image edges, is used to indicate if further actuation of the motor, or otherwise, is required. The method is compared to conventional statistical methods (standard deviation and kurtosis). Results indicate that the MoZAK method presents a viable distance estimator compared to conventional statistical methods.