Aerial image prediction

Computer Vision (CV) is a Hot field of study in Computer since, this field deals with automatic processing of images in order to extract relevant information from them in order to draw conclusions and make informed decisions.
CV provide wide variety of apps. One of the aspects studied and developed in CV is navigation and tracing. In this work we developed an algorithm that improves the tracing ability of a flying object.

Our algorithm predicts the next frame of a video stream taken from a flying object camera  by using four major steps:

- Calculating flying object speed and direction.

- Use maximum data from the last frame.

- Search a suitable data in all previous frame to complete missing parts.

- Stich the suitable data with no seams.     

To do so we integrated several ideas from previous papers:

1) Motion detection using Optical flow algorithm.

2) Scene completion using millions of photographs, James Hays, Alexei A. Efros 2007

3) Poisson Image Editing, Perez 2003

Using all 3 papers to build our algorithm required a vast literary exploration, finding the right parameters and a lot of experimentation.

In this project our goals were:

- Explore the possibility of prediction of future frame in a video stream taken from a flying camera, when using only previous frames of the stream.

- Provide an algorithm that predicts the next frame with maximum accuracy.

The algorithm managed to predict the images with satisfactory results.