Depth Estimation Using Single Fish-Eye Camera

Abstract

Nowadays, with the increasing application of navigation systems, robotic systems especially for unmanned platforms, omnidirectional measurement technology is in significant demand. The omnidirectional distance measurement plays an indispensable role in path planning and object avoidance. The recent studies for distance measurement are focused on laser or ultrasonic distance finding, yet they are considered to own high cost, high computational burden and low accuracy. Specifically, the studies for analytical solutions are more effective but insufficient due to visual servoing. Under this situation, one analytical theory research of visual servoing for omnidirectional distance measurement is essential and requires to be considered thoroughly.

To generate the most suitable omnidirectional distance measurement catering for the specific task requirement, the distance measure sensor should be optimized with considering the effective information from the environment. The one that can be mentioned here is that the conventional researches only emphasize the accuracy for the measurement and neglect the field of view (FOV) and computational burden. So the emergence of the low-efficacy phenomenon for robot control is inevitable. It is necessary to propose and develop one novel method to optimize the omnidirectional distance measurement method for specific tasks.

The research will present the visual servoing method for the robot control catering to the specific task. Firstly, the analytical problem for traditional measurement is are illustrated and analyzed. Secondly, visual detection method such as object tracking, detection and human pose detection method to collect the image size is illustrated.  Thirdly, the image correction method and camera calibration method are involved in optimizing the size and distance measurement to get the best performance in executing the desired tasks. Finial the system will be established and employed to verify the possible findings in the future.