Death by Kinect

The purpose of the project is to visually demonstrate in a real-time system the performance gains of leveraging the parallel processing of a many-core architecture. The system uses the performance of computation intensive computer vision algorithms to measure the speed-up between a single-core and a parallelized implementation. The project performs a comprehensive analysis of the parallelism of automatic target recognition and object tracking algorithms running on a many-core system.There are two main objectives for this project. One objective of the project is to be able to parallelize a portion of a program on a GPU (NVidia CUDA) to see an improvement in the amount of data that is processed in the same amount of time as a result of the parallelization. Another objective is to implement two autonomous systems to use as the basis for the parallelization tests. Each system will be built using already existing technologies. One system will be a target drone that will be avoiding a projectile that is shot at it. The other system will be a stationary robot that will recognize the target drone and fire a projectile at it. The novelty of the project comes from combining disparate technologies for the purpose visually demonstrating the effects of parallelization.The optimization strategy to leverage the many-core architecture is to replace time-sensitive functions. The functions identified relate to the object detection and tracking. The OpenCV library is used to provide both capabilities. Fundamental matrix calculation functions will also be replaced with massively parallel versions. These optimizations will have a direct impact on the Missile Turret Platform hitting the Target Drone and the Target Drone System evading the projectile threat.