Vision-based object distance measurement using mono and stereo vision

Abstract EN

Visual servoing (VS) is an essential research field because it is included in several robots' applications, such as "navigation and visual surveillance".

The VS comprises three stages: taking an image by a visual sensor, processing the image to detect a target object, and guiding the robot to navigate the object.

In this work, this issue will be studied to find the object's position in the frame of the image; and then to estimate its distance from the robot frame.

The paper will provide the fundamental information of using two techniques: mono and stereo vision.

C++ programming in addition to OpenCV libraries were used to develop and implement all codes.

The experimental results demonstrated that the distances measured by stereo-vision were relatively accurate with errors ranged from -2.0 to 2.6 % in comparison with mono-vision's measurement errors that were fluctuated between -6.4 and 2.5%.

It was also verified that the better results were achieved by using the larger distance between the cameras within the stereo-vision system.