Method for correcting abnormal point cloud

Patent metadata

Field	Value
Publication number	US-11875477-B2
Application number	US-202117354188-A
Country	US
Kind code	B2
Filing date	Jun 22, 2021
Priority date	Dec 1, 2020
Publication date	Jan 16, 2024
Grant date	Jan 16, 2024

Abstract

A method for correcting abnormal point cloud is disclosed. Firstly, receiving a Primitive Point Cloud Data set by an operation unit for dividing a point cloud array into a plurality of sub-point cloud sets and obtaining a plurality of corresponding distribution feature data according to an original vector data of the Primitive Point Cloud Data set. Furthermore, recognizing the sub-point cloud sets according to the corresponding distribution feature data for correcting recognized abnormal point cloud. Thus, when the point cloud array is rendered to a corresponding image, the color defect of the point cloud array will be improved or decreased for obtaining lossless of the corresponding image.

First claim

What is claimed is:

1. A method for correcting abnormal Point Cloud Data, applied for a Correction System including an Image Processing Unit and an Optical Scanning Unit, the Optical Scanning Unit scanning an outer structure to create a Primitive Point Cloud Data and transmit the Primitive Point Cloud Data to a Data Base, the method for correcting the abnormal Point Cloud Data comprising the steps of:
the Image Processing Unit reading the Primitive Point Cloud Data from the Data Base; the Primitive Point Cloud Data includes a plurality of Normal Point Data and at least an Error Point Data and a Primitive Voxel Space Data;
the Image Processing Unit following the Primitive Voxel Space Data to divide the Primitive Point Cloud Data and obtain a plurality of Sub-Point Cloud Sets and the corresponded plural Distribution Feature Data; these Sub-Point Cloud Sets have the Normal Point Data and the Error Point Data;
the Image Processing Unit following the Distribution Feature Data to identify the Sub-point Cloud Sets and obtain the Error Point Data; and
the Image Processing Unit following a Distribution Feature Data corresponding to the Error Point Data to make regression operation and correct the Error Point Data into a Normal Point Data.

2. The method for correcting the abnormal Point Cloud Data of claim 1 , wherein the Normal Point Data and the Error Point Data include a coordinate data, a color data and an intensity data; the coordinate data corresponds to the Primitive Voxel Space Data.

3. The method for correcting the abnormal Point Cloud Data of claim 2 , wherein the Distribution Feature Data include a plurality of position eigenvalues, color eigenvalues and intensity eigenvalues corresponding to the Normal Point Data and the Error Point Data.

4. The method for correcting the abnormal Point Cloud Data of claim 1 , further comprising the steps of:
using the Optical Scanning Unit to point-by-point form image and create the Point Cloud Data.

5. The method for correcting the abnormal Point Cloud Data of claim 4 , wherein the Optical Scanning Unit is a Lidar, a 3-D laser scanner or a light-beam scanner.

6. The method for correcting the abnormal Point Cloud Data of claim 1 , wherein in the steps that the Image Processing Unit follows the Primitive Voxel Space Data to divide the Primitive Point Cloud Data and obtain a plurality of Sub-Point Cloud Set and the corresponded plural Distribution Feature Data, the steps include:
the Image Processing Unit following the plural Voxel Grids of the Primitive Voxel Space Data to divide the Primitive Point Cloud Data into the Sub-Point Cloud Sets;
the Image Processing Unit performing a nearest neighbor index operation, a principal component analysis operation and a de-constraint conversion operation according to the Primitive Point Cloud Setting and obtaining the Distribution Feature Data, which respectively correspond to all neighboring points of each image point in the Sub-Point Cloud Sets;
the Image Processing Unit categorizing the Sub-Point Cloud Sets according to the Distribution Feature Data to obtain a plurality of category labels; and
the Image Processing Unit labels the Sub-Point Cloud Sets according to the category labels.

7. The method for correcting the abnormal Point Cloud Data of claim 6 , wherein the Image Processing Unit performs the nearest neighbor index operation by running a K-Nearest Neighbor (KNN) search algorithm to run the Primitive Point Cloud Data; the Image Processing Unit performs the principal component analysis operation and the conversion operation according to the three-axis variances and the Primitive Point Cloud Data; the Image Processing Unit performs the de-constraint conversion operation by means of logarithmic operation according to the three-axis variances, in the prospective of removing the boundaries corresponding to these variances.

8. The method for correcting the abnormal Point Cloud Data of claim 1 , wherein in the step that the Image Processing Unit follows the Distribution Feature Data to identify the Sub-point Cloud Sets, the Image Processing Unit follows the categories of the Sub-point Cloud Sets and the corresponded plural label numbers to identify the Normal Point Data and the Error Point Data.

9. The method for correcting the abnormal Point Cloud Data of claim 1 , wherein in the steps that the Image Processing Unit follows the Distribution Feature Data to identify the Sub-point Cloud Sets, the Image Processing Unit follows at least a second category cloud of the Sub-point Cloud Set and at least a corresponded label number to identify and obtain the Error Point Data.

10. The method for correcting the abnormal Point Cloud Data of claim 1 , wherein in the steps that the Image Processing Unit follows the corresponded Distribution Feature Data to correct the Error Point Data, the steps include:
the Image Processing Unit following the Error Point Data and the color data of at least a Neighboring Point Data to perform regression operation and obtain a First Color Correction Data corresponding to the Error Point Data;
the Image Processing Unit following the color data of the Neighboring Point Data of the First Color Correction Data and the Error Point Data to perform regression operation and obtain a Second Color Correction Data corresponding to the Error Point Data; and
the Image Processing Unit following a weighted average method for the Sub-Point Cloud Set and combines it with the First Color Correction Data and the Second Color Correction Data to obtain a Standard Color Correction Data, used to correct the Error Point Data.

11. The method for correcting the abnormal Point Cloud Data of claim 1 , wherein in the steps that the Image Processing Unit follows the corresponded Distribution Feature Data to correct the Error Point Data, the steps include:
the Image Processing Unit following a position data corresponding to the Error Point Data to perform regression operation and obtain a Position Regression Data;
the Image Processing Unit following an Image Capture Data corresponding to the Error Point Data to read an image color data and perform regression operation to obtain a Color Regression Data; and
the Image Processing Unit following a weighted average method corresponding to the Sub-Point Cloud Sets and combining it with the Position Regression Data and the Color Regression Data to build a Color Regression Unit and correct the Error Point Data.

Assignees

• Nat Applied Res Laboratories

Inventors

• Wang Chih-Wei
• Lai Chuan-Lin
• Kuo Chia-Chen
• Wu I-Chen

Classifications

• G06T7/0002 Primary

  Inspection of images, e.g. flaw detection · CPC title

• G06T5/00

  Image enhancement or restoration · CPC title

• G06T5/001 Primary

  Physics · mapped topic

• G06F18/2135

  based on approximation criteria, e.g. principal component analysis · CPC title

• G06T2207/10024

  Color image · CPC title