This ongoing project pursues the development of an on-line monitoring system allowing to reliably extract certain features (reference points) and new popping-up objects, that are needed to be related to the reference points, based on two streams of video data – EO and IR. Specifically, the objectives are as follows:
- estimate capability of processing visual / IR video data in MATLAB/Simulink development environment;
- find surveyed reference points on the visual / IR images;
- try to find multiple popped-up objects within the visual / IR image;
- combine visual and IR images together to improve the robustness of the video scoring algorithms;
- assess the developed tools for the real-time usage (blending EO and IR streams together).
To date, several algorithms / tools were developed to address these objectives. These algorithms / tools were devoted to address a sequence of particular problems:
- loading the sequence of the images into MATLAB development environment and creating a convenient analyzing tool allowing to determine the range of the frames that might be of a specific interest;
- finding a way to “normalize” consequent images, which provides a capability to use their difference as a reliable tool for new objects detection;
- finding multiple pairs of scale-invariant feature points to find the corresponding transform between two images;
- detecting all surveyed reference points (and newly popped-up points) on each image.
The developed software has exhibited an excellent performance and assumes further development / tuning / field testing.