Work developed for the Image Processing and Vision course from Instituto Superior Técnico.

The goal of the project is the detection, localization and tracking of moving objects using information from depth cameras. The cameras are in a fixed position and the scene is composed of a static set of objects (background) and a variable number of moving objects.

• Considering just one camera for each detected object in each frame a box must be draw around it.
• Output: for every tracked object a struct called objects is returned, with 4 fields the first one being the X coordinates of the box, the second the Y coordinates of the box, the third the Z coordinates of the box and a fourth one called frames_tracked containg the frames where the objects were detected.

• Considering two cameras, for each detected object (no matter the camera) in each frame a box containing the identified object must be draw.
• Output: same as the part 1 plus another struct called cam2toW containing the rotacional and translation matrices of the camera 2 in relation to the world coordinate system(camera 1).

For part 1 use the files in the images1 or images4 folders. Only change the name of the folder on the top of the file part1_test. For part 2 use the files in the images2 or images3(never tested) folders. Pay attencion to change the name of the folder in the top of the file part2_test.

VLFeat library is required: VLFeat library

• Luís Lopes
• Diogo Morgado
• Valter Piedade