Target Detection using Image Processing Techniques

摘要

This paper describes the Target Detection techniques used for building an Autonomous UAV. This UAV is designed for participation in Ohio Autonomous Aerial Vehicle Competition- 2014 (AAVC) in which the Quadcopter is required to navigate around the obstacles in GPS denied and hazard cluttered indoor environment, detect the target along with its location uncertainty and wirelessly transmit the image of the target to the ground station. The target is detected using image processing algorithms written in Python. The algorithm uses the blob and shape detection on Hue, Saturation and Value (HSV) color space to ensure detection under various conditions of lighting, shadows, and distance. The algorithm is then tested in a variety of scenarios that account for variance in lighting and shape detection. Challenges are limited computational processing on board, varying ambient light conditions, and detection of correct target in the presence of more than one same colored targets. To increase the robustness of this method we assume there maybe multiple targets of the same color; however these objects will be of somewhat different shapes. We are also assuming no object will be blocking the onboard cameras field of view of the target at the entrance of the target area.