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 [2017/02/03 00:19] – [Module #2: Point Cloud Algorithms] 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 ======
  
-3D point cloud is a commonly used data in spatial modeling & understanding. Such a data can be captured from laser scanners, stereo camera, or structured-light cameras. Point cloud processing is essential for applications like self-driving cars, infrastructure inspection, construction modeling, etc.. Here we will go through some basic modules of a point cloud processing pipeline, include 1) clustering, 2) normal vector estimation, 3) point cloud descriptor extraction, 4) point cloud registration and 5) triangulation.+3D point cloud is a commonly used data in spatial modeling & understanding. Such a data can be captured from laser scanners, stereo camera, or structured-light cameras. Point cloud processing is essential for applications like self-driving cars, infrastructure inspection, construction modeling, etc.. 
  
 ===== Module #1: A Basic Point Cloud Processing System  ===== ===== Module #1: A Basic Point Cloud Processing System  =====
 +----------------------------------
  
-__** Week #1: 3D Points Class and Basics**__ \\ +Here we will go through some basic modules of a point cloud processing pipeline, include 1) clustering; 2) normal vector estimation; and 3) triangulation.
-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 class stub (TO BE PROVIDED ON FRIDAY OR SATURDAY) and the sample point cloud file here (TO BE DONE).  Perform the following: +  - [[ECE4580:Module_PCD:Basics | Week 1]]: Point Cloud Basics. 
-  - Flesh out the **plot** function so it can plot the data. +  - [[ECE4580:Module_PCD:Connected | Week 2]]: Proximity and Connected Components. 
-  - Flesh out the **normalizePose** function. +  - [[ECE4580:Module_PCD:Normals | Week 3]]: Local Surface Normal Estimation. 
-  - Flesh out the **numPoints** function (look into the pointCloud class to see how to obtain the value). +  - [[ECE4580:Module_PCD:ClusterNormal | Week 4]]: Clustering via Proximity and Local Normal Alignment. 
-  - Flesh out the **removeInds** function (again look into the pointCloud class operations to see how to make this easier). +  - [[ECE4580:Module_PCD:Triangulation | Week 5]]: A Simple Triangulation Scheme. 
- +  - [[ECE4580:Module_PCD:TriangulationApply | Week 6]]: A Simple Approach for Point Clouds to Surfaces. 
-Regarding the pose normalization function, here are some [[ECE4580:Module_PCD:NormalizePose|notes]]. +  - [[ECE4580:Module_PCD:BridgeProblem | Week 7]]: Bridge Visualization and Comparison. Do Part #1. 
- +  - [[ECE4580:Module_PCD:BridgeProblem | Week 8]]: Bridge Comparison. Do Part #2.
-/*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 connected component clustering.  This will involve: +
- +
-  - Completing the connectivity matrix member function. +
-  - Parsing the function to find [[ECE4580:Module_PCD:ConnectedComponents|connected components]]. +
-  - Giving each component a different color for plotting purposes. +
-  - Returning the label of each point in a vector list. +
- +
-Be careful if you try to run this on random 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). +
- +
-__** 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 local normal estimation.  Let's try out both. +
- +
-  - One is via principal component analysis, otherwise known as [[ECE4580:Module_PCD:NormalPCA|PCA]]. +
-  - The other is using the [[ECE4580:Module_PCD:NormalSVD|singular value decomposition]] we know and love (by now). +
- +
-__** 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 **__ \\ +
-Read the page of Delaunay triangulation on [[https://en.wikipedia.org/wiki/Delaunay_triangulation|Wikipedia]]. For further understanding, refer to the slides from [[https://people.csail.mit.edu/indyk/6.838-old/handouts/lec9.pdf|MIT CSAIL]].   +
- +
-NEED TO FLESH OUT MORE. +
- +
-/* +
-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. +
-*/+
  
 ===== Module #2: Point Cloud Algorithms ===== ===== Module #2: Point Cloud Algorithms =====
 +---------------------------------------------
 +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 91: Line 39:
 */ */
  
-===== Module #3: Object Recognition with Point Clouds =====+/*__** Week #X: Cloud Proximity to Another Cloud
  
-TO BE WRITTEN.+Get set of points that a given point cloud is near to relative to another point cloud. 
 + 
 +//Notes:// Matlab function: ''pdist2'' 
 +*/
  
 -------- --------
ece4580/module_pcd.1486099188.txt.gz · Last modified: 2024/08/20 21:38 (external edit)