The Israeli navy faces daily attempts of breaching Israel's teritorial waters. This project aims to develop an algorithm to be installed on cutting edge equipment that will identify targets
Abstract
The Israeli navy faces daily attempts of breaching Israel’s teritorial waters. This project aims to develop an algorithm to be installed on cutting edge equipment that will identify targets in a naval environment. As first stage, single frames are analyzed in order to find the horizon in the frame. This important goal provides the ability of detecting moving objects, by comparing consecutive frames of a given video. Assuming that only objects are moving between frames (the environment is moving much slower), they will be detected using MSE comparison between the frames. Thus, working on each frame of the video input, we can detect and mark all moving objects.
The problem (or the background)
The main problem to be solved in this project is identifying moving targets in a naval evironment. This algorithm can be used a part of a tracking system installed on USV’s (Unmanned Surface Vehicles) and will provide analysis of the input video. The problem of detecting objects in a naval environment is not easy and many assumptions are to be taken (characteristcs of the input video, the objects, the platform etc). This project may serve as basis to a more general algorithm which will provide more robustness.
The solution (or the basic approach)
The solution implemented at this algorithm includes the following steps: 1. Dividing a single frame into four quarters (the original input video is soldered from four cameras) 2. Finding the horizon in a single quarter according to the following parameters: 1. Constance force. 2. Derivative in Y axis only. 3. Local and global histogram. 3. Averaging the horizon incline of three consecutive frames at the proper quarter And rotating the quarter achieving zero angle of horizon comparing to X axis. 4. Averaging the horizon of Y value of three consecutive frames at the proper quarter And shifting the quarter to the average value. 5. averaging the horizon of Y value of the four quarters soldering a single frame straighten the whole Frame to the average value- Horizon found!! 6. Matching two consecutive frames according to MMSE algorithm and creating a diff picture of them. 7. Matching two consecutive diff pictures in order to emphasize the diff. The diff found represent moving object at the frame, therefore may imply a suspicious object.
