Patricio Vela: Course Wiki

User Tools

Site Tools

Trace:
ece4580:module_pcd

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revisionPrevious revision
Next revision		Previous revision
ece4580:module_pcd [2017/04/08 17:38] – pvelaece4580:module_pcd [2024/08/20 21:38] (current) – external edit 127.0.0.1
Line 1: Line 1:
 +~~NOTOC~~
 ====== Point Clouds Processing and Interpretation ====== ====== Point Clouds Processing and Interpretation ======
  
Line 14: Line 15:
   - [[ECE4580:Module_PCD:Triangulation | Week 5]]: A Simple Triangulation Scheme.   - [[ECE4580:Module_PCD:Triangulation | Week 5]]: A Simple Triangulation Scheme.
   - [[ECE4580:Module_PCD:TriangulationApply | Week 6]]: A Simple Approach for Point Clouds to Surfaces.   - [[ECE4580:Module_PCD:TriangulationApply | Week 6]]: A Simple Approach for Point Clouds to Surfaces.
-  - [[ECE4580:Module_PCD:BridgeProblem | Week 7]]: Bridge Visualization and Comparison. +  - [[ECE4580:Module_PCD:BridgeProblem | Week 7]]: Bridge Visualization and Comparison. Do Part #1. 
- +  - [[ECE4580:Module_PCD:BridgeProblem | Week 8]]: Bridge Comparison. Do Part #2.
-==== Week #1: 3D Points Class and Basics ==== +
-For this week, we will start with some basis stuff. There is a basic point cloud stub class that we will use to augment the existing Matlab [[https://www.mathworks.com/help/vision/ref/pointcloud-class.html|point cloud class]] and [[https://www.mathworks.com/help/vision/3-d-point-cloud-processing.html|toolbox]] for point cloud processing.  Note that the point cloud capability is in Matlab R2015b and later. +
- +
-Download the **pts3D** Matlab  [[http://pvela.gatech.edu/classes/files/ECE4580/Module_PCD/pts3D.m|class stub ]] and the sample Matlab file [[http://pvela.gatech.edu/classes/files/ECE4580/Module_PCD/points01.mat|here]]. The Matlab file has four point matrices in it. +
- +
-Perform the following: +
-  - Flesh out the **plot** function so it can plot the data (plot the different point matrices). +
-  - Flesh out the **normalizePose** function (normalize for the sPts0X point matrices). +
-  - Flesh out the **size** function (look into the pointCloud class to see how to obtain the value). +
-  - Flesh out the **removeInds** function (again look into the pointCloud class operations to see how to make this easier). +
-  - Flesh out the **keepOnlyInds** function. +
- +
-//Notes:// Regarding the pose normalization function, here is the [[ECE4580:Module_PCD:NormalizePose|math description]]. You should use the ''pointCloud'' class member functions as much as possible. Also the function ''setdiff'' is your friend. +
- +
-//Explore & Deliverables:// Given what you know about the SVD and how it orders the singular values, relate this to the shapes of the different ''sPts0X'' variables.  Write in English what you think is happened and what role the SVD plays in doing that.  Plot the before and after pose normalization for the ''sPts0X'' point sets. +
- +
-When you load the points into a matrix, use the ''size'' or ''numPoints'' member functions to list the number of points in the point cloud, for each point matrix provided.  Show that you can remove points from a point cloud by plotting the original point cloud and a decimated version obtained by removing more than one third of the points using the ''removeInds'' function. Display the number of points before and after the removal operation. +
- +
- +
-/*Grab data sets, learn to load point cloud data and visualize it.  +
- +
-A list of point cloud dataset can be found [[http://www.pointclouds.org/news/2013/01/07/point-cloud-data-sets|here]]. +
- +
- +
-For those feel comfortable with C++ programming, [[http://pointclouds.org/documentation/tutorials/#i-o|Point Cloud Library (PCL)]] is a nice alternative option. +
-*/ +
- +
-==== Week #2: Connected Component Clustering ==== +
-Flesh out the member functions for performing [[ECE4580:Module_PCD:ConnectedComponents|connected components clustering]] (''clusterByProximity'' and ''connectivityMatrix'').  This will involve: +
- +
-  - Completing the connectivity matrix member function. +
-  - Parsing the matrix to find connected components. +
-  - Giving each component a different color for plotting purposes. +
-  - Returning the label of each point in a vector list. +
- +
-Be careful when you try to run this on arbitrary data.  Because the connected components matrix is memory intensive, it can only be done with point clouds that are relatively small (on the order of 20 thousand points). For more than that many points, the best option is to utilize Matlab's internal pointCloud member functions to request the local neighborhood set. +
- +
-//Notes:// Matlab functions that will be useful include: ''pdist'', ''squareform''. You will also need to figure out how to implement the algorithm through judicious use of temporary variables. +
- +
-//Explore & Deliverables:// Apply to the set of points from the Week #1 data file. The data file clearly has distinct connected components.  Show that your code does grab them properly. +
- +
-==== Week #3: Local Normal Vector Estimation ==== +
-If the point cloud data represents an object as scanned from some device, usually the points lie on the surface of the object. So they are really 2D structurally speaking (local to each point), though they are in 3D.  Let's write a function to estimate the local normals to the point cloud, plus one more to visualize them. There are actually two ways to perform the [[ECE4580:Module_PCD:LocalNormals|local normal estimation]].  Let's try out both. +
- +
-  - One is via principal component analysis, otherwise known as [[ECE4580:Module_PCD:LocalNormals#PCA|PCA]]. +
-  - The other is using the [[ECE4580:Module_PCD:LocalNormals#SVD|SVD]] we know and love (by now). +
- +
-For the sample point cloud file given, plot the normals. Do not really plot them all, but rather sub-sample the points array and only compute then plot the normals for those points. The Matlab example linked to in the discussion page for this problem (above link for //local normal estimation//) shows how to perform that. +
- +
-==== Week #4: Clustering by Normal Structure ==== +
-The connected component clustering does not do too well when different objects are actually connected, say because they are touching, or one is resting on the other.  Usually, however, the objects might have differing local normals at the interface where they are touching.  Let's add normal agreement to the connected component clustering to see how the normals can be used to differentiate parts of the scene. +
- +
-Perform a region growing clustering strategy using the connected components for the region expansion candidate, and using the angular difference between normals to assess membership to the same cluster. +
- +
-/*Read the slides on [[http://www.connellybarnes.com/work/class/2013/shape/04_normal.pptx|normal vector estimation]]. +
- +
-For Matlab folks, an exemplar implementation of normal vector estimation can be found [[https://vision.princeton.edu/pvt/points2normals.m|here]]. Try to implement your own estimator. +
- +
-For PCL folks, you may refer to the [[http://pointclouds.org/documentation/tutorials/normal_estimation.php|sample code]]. Try to implement your own estimator. +
-*/ +
- +
-==== Week #5: Triangulation ==== +
- +
-When the point cloud represents a surface, then it is possible to create a triangulated mesh for more rapidly visualizing the object.  Plotting points clouds takes some non-trivial processing since the renderer has to plot many points and do so with occlusion handling and scale awareness.  A must faster means to visualize objects whose surface are modelled by the point cloud would be to simply the points to surfaces, then to plot those surfaces.  The main idea is that there will be far less surfaces to plot than there are points.  Furthermore occlusion handling, and possible even scale handling, is much easier with surfaces. +
- +
- +
- +
-Given an unordered point cloud, the process of generating surface patches that attempt to reproduce the actual surface is known as //mesh triangulation// or //surface triangulation//.  For planar point clouds or volumetric point clouds (cannot be planar), it is possible to triangulate or tesselate the surface using  +
-[[https://en.wikipedia.org/wiki/Delaunay_triangulation|Delaunay triangulation]]. For further understanding, refer to the slides from [[https://people.csail.mit.edu/indyk/6.838-old/handouts/lec9.pdf|MIT CSAIL]].   +
-These methods work with convex hull type processing, so they don't quite do the trick. +
- +
-Here, you will explore an alternative strategy to surface mesh triangulation, which is a naive version of the //advancing fronts// method. [[ECE4580:Module_PCD:Triangulation01| Read on here]]. +
- +
-//Explore and Deliverables:// In addition to the surface triangulations for the sample point clouds and one object from the earlier point cloud, discuss the performance of the method as the main parameter ($\rho_{near}$) is adjusted.  What happens?  Is it sensible? Can you relate its value to what you see happening? Can you relate it to concepts you might know from signal processing or some ECE class you've taken? +
- +
-/* +
-For Matlab folks, an exemplar implementation of Delaunay triangulation can be found [[https://www.mathworks.com/help/matlab/ref/delaunay.html|here]].  +
- +
-For PCL folks, there's no sample code you can use this time :( However you may check the C++ implementation of Delaunay triangulation from [[https://github.com/Bl4ckb0ne/delaunay-triangulation|here]]. Try to integrate it into your PCL pipeline. +
-*/ +
- +
-==== Week #6: Triangulation on Objects ==== +
- +
-Combine the clustering code with the triangulation code to come up with a cluster-driven surface mesh of objects in a point cloud. The idea is that, first the point cloud is clustered based on proximity and normal agreement.  Each of cluster is extracted and triangulated. All of the meshes get put together for connected components (ignoring the normal agreement). If possible, color each mesh. Plot the final output.  It should be a set of meshes for the distinct/disconnected objects in the scene. If everything is one big connected mess, then that's a problem.+
  
 ===== Module #2: Point Cloud Algorithms ===== ===== Module #2: Point Cloud Algorithms =====
---------------------------------- +--------------------------------------------- 
-Here we cover 3) point cloud descriptor extraction, 4) point cloud registration+Here we cover 1) point cloud registration and 2) point cloud descriptor extraction.
  
-__** Week #1: Point Cloud Registration **__ \\+  - [[ECE4580:Module_PCD:ICP | Week 1]]: Aligning matching point clouds 
 +  - [[ECE4580:Module_PCD:Moments | Week 2]]: A Basic point cloud descriptor 
 +  - [[ECE4580:Module_PCD:Descriptors | Week 3]]: Local descriptors.
  
-A commonly used and somewhat simple method for registering two point clouds (that presumably are of the same object or have significant similar structure), is to use what is called Iterative Closest Point (ICP). 
-Read up on [[http://cs.gmu.edu/~kosecka/cs685/cs685-icp.pdf|ICP]], then implement in Matlab. 
  
-MORE DESCRIPTION NEEDED HERE. +===== Ignore =====
- +
-/* +
-For Matlab folks, an exemplar implementation of ICP can be found [[https://www.mathworks.com/help/vision/examples/3-d-point-cloud-registration-and-stitching.html|here]]. Try to implement your own version of ICP. +
- +
-For PCL folks, you may refer to the [[http://pointclouds.org/documentation/tutorials/iterative_closest_point.php#iterative-closest-point|sample code]]. Try to implement your own version of ICP. +
-*/+
  
 __** Week #X: Point Cloud Segmentation **__ \\ __** Week #X: Point Cloud Segmentation **__ \\
Line 128: Line 39:
 */ */
  
-__** Week #X: Cloud Proximity to Another Cloud+/*__** Week #X: Cloud Proximity to Another Cloud
  
 Get set of points that a given point cloud is near to relative to another point cloud. Get set of points that a given point cloud is near to relative to another point cloud.
  
 //Notes:// Matlab function: ''pdist2'' //Notes:// Matlab function: ''pdist2''
- +*/
-===== Module #3: Object Recognition with Point Clouds ===== +
------------------------------------ +
-TO BE WRITTEN.+
  
 -------- --------
ece4580/module_pcd.1491687502.txt.gz · Last modified: 2024/08/20 21:38 (external edit)