Week - 5 3D Tracking Continued...

This week we will be focusing on creating an STMap to undistort images, 3D tracking using CameraTracker.

First we load the footage to the node which is not undistorted, we will check the properties of the footage to check what type of lens, which camera and resolution type which was used to shoot the footage. After knowing the type of lens used we downloaded the checkered image of the lens type and
 reformated the image, added a grade node to establish the white and black points, added the LensDistort node - estimates the lens distortion in a given image, either through Grid Detection or manual Line Detection. We now go to the properties of the lensdistort and in analysis we try to detect the lines. After Detecting we try to add, remove lines and make them as straight as possibles using points from one end to another end from vertically to horizontal keeping in mind that each vertical and horizontal planes contain a single straight line or else Nuke tries to Straighten the line which maybe connected to no straight lines.


After completing the above step, we solve the image.

Go to the lens distortion node properties, under the lens distortion menu, output, change the output mode to STMap, then add a write node, select all in channels and save it as .exr file and change the datatype to 32-bit and then render the image, now we have created an STmap which can be used for any project having the same camera lens.

Now bring the .exr file to the project and the image sequence, now we use a STmap node - node allows you to move pixels around in an image. Connect the src to the image sequence and stmap to the STMap exr file add the write node and save the file as a .jpeg file and render the sequence now we have an image sequence that is now been Undistorted.

In the afternoon session we focused on 3D tracking using Undistorted Image, we brought in the undistorted image sequence, added a cameratracker node and connected it to the viewer. Coming to the cameratracker properties we can give more info to nuke to increase the efficiency of camera track, we proceed to the settings section and increase the no. of features and select the preview the features. Now we track the sequence.


After tracking we simply solve the tracks, now we have points which are green, orange and red colours. The Green colour tells us that the track is good, Red colour tells us that the track is bad and the Orange colour tells us that the track was not solved. We now try to bring down the Solve Error by going to the autotracks and playing with the settings and trying to bring the SolveError to less that 1.0. After bringing it down to less that 1.0 we export it to a Scene and create it.

Next, add a ScanlineRender node. This takes our 3D elements and renders them back into a 2D image. Now we try to set the ground plane for reference by selecting the points near to the graoung and grouping to the ground plane. When viewed in the 3D space we get this below image,


Comments

Popular posts from this blog

Final_Project

Week - 4 3D Workspace in Nuke

Week - 3 2D Tracking: Tracker, Planar Tracker, Rotopaint, Match-moving