Background estimation

4 novembro 2019, 17:00 • João Paulo Salgado Arriscado Costeira

Given a sequence taken from a static camera, how to compute the background image using the median (explained in class). Check with twocams_books dataset.

imgmed=zeros(480,640,45);

for i=1:45,

    load(['depth1_' int2str(i) '.mat']);

    imagesc(depth_array);

    imgmed(:,:,i)=double(depth_array)/1000;

    pause(.1)

    drawnow;

end

bg=median(imgmed,3);

d=dir('depth1*.mat');

for i=1:length(d),

    load(d(i).name);

    imagesc(abs(double(depth_array)/1000-bg)>.25);

    colormap(gray);

    pause(1);

end

IV - Image Features (Scale and SIFTs)

4 novembro 2019, 15:30 • José Santos-Victor

Image scale. Invariance to scale and scale detection.
Difference of Gaussians as scale estimation.
The SIFT feature detector.
Point detection and region representation through gradient orientation histograms.

3D point cloud registration and feature matching

31 outubro 2019, 14:00 • João Paulo Salgado Arriscado Costeira

Objective:

1.      How to register 2 point clouds

2.      Procrustes problem

3.      How to match 2 images and compute 3D transformation using depth info

3D point cloud registration and feature matching

31 outubro 2019, 11:00 • João Paulo Salgado Arriscado Costeira

Objective:

1.      How to register 2 point clouds

2.      Procrustes problem

3.      How to match 2 images and compute 3D transformation using depth info

3D point cloud registration and feature matching

31 outubro 2019, 08:00 • João Paulo Salgado Arriscado Costeira

Objective:

1.      How to register 2 point clouds

2.      Procrustes problem

3.      How to match 2 images and compute 3D transformation using depth info