Authors

Presenter(s)

Arjun Udayakumar Sherly

Files

Download

Download Full Text (890 KB)

Description

Rear end traffic crashes are rising alarmingly but these crashes can be avoided effectively with the help of an efficient rear end camera system that could predict the direction of the oncoming vehicle and change the navigation of parent vehicle appropriately. The proposed approach tackles the problem by providing a camera system at the rear end of the vehicle that notifies the vehicle about the acceleration and direction of the approaching object and notifies the action to be taken so as to avoid the threat or collision by employing the principle of optic flow. The optic flow of the object of interest is calculated and tracked for extracting the key features. The three prominent features that are evaluated closely are acceleration, size and the direction of the object of interest. Acceleration determines how fast the object is moving towards the rear station. Direction attribute determines whether the object is moving towards or away from the rear station. Modified Kanade Lucas Tomasi Algorithm (KLT) calculates the optical flow of object of interest. The stationary objects such as trees, street lights and buildings are ignored as the background. The modified KLT uses a Pyramidal approach to evaluate the severity of threats. Pyramidal approach at different levels takes care of objects moving at high speed that might disappear from the frame. The testing and evaluation is done in Husky robot from Clear Path Robotics and using Robot Mobility Platform 220 by Segway. In the era of exponentially growing trend in autonomy, Rear Eye Vision stands as the epitome of research in the field of autonomous navigation.

Publication Date

4-5-2017

Project Designation

Graduate Research - Graduate

Primary Advisor

Vijayan K Asari

Primary Advisor's Department

Electrical and Computer Engineering

Keywords

Stander Symposium poster

Rear Eye Vision for Enhanced Safety in Autonomous Navigation

Share

COinS