Point Cloud Resampling Using Graph Signal Processing

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for simplifying point clouds often result in data loss or require significant computational resources, as they either risk losing key features or are complex and time-consuming.

Innovation Solution

The approach involves resampling point clouds to preserve a subset of key points, using graph signal processing to select points based on specific applications, with randomized strategies that optimize reconstruction error and preserve information efficiently.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If sub-sampling decimation or uniform spatial clustering is used to reduce point cloud size, then the data volume is reduced, but key features of objects and surfaces are lost

Engineering Contradiction:
Improvepoint cloud data volumeVSAvoidkey features of objects and surfaces
Core Design Contradiction:
Quantity of substanceVSLoss of information

Solution Approach 1:

The patent applies local quality by differentiating between important and unimportant regions in the point cloud. It computes a significance metric for each point based on local geometric properties (e.g., curvature, density) and selectively preserves points in high-significance regions while removing points in low-significance regions. This ensures that key features such as edges, corners, and surface details are maintained while reducing overall data volume.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent changes parameters by transforming the point cloud representation from a uniform set of points to a selectively sampled set based on computed significance metrics. It adjusts the sampling density dynamically according to local geometric characteristics, using parameters such as point density thresholds, curvature thresholds, and significance scores to control the reduction process while preserving essential features.

Inventive Principle:
Principle #35Parameter changes

2Loss of information

If complex simplification methods are used to preserve key features, then data quality is maintained, but computational expense and processing time increase

Engineering Contradiction:
Improvekey features preservationVSAvoidcomputational expense and processing time
Core Design Contradiction:
Loss of informationVSDevice complexity

Solution Approach 1:

The patent applies preliminary action by pre-computing geometric properties and significance metrics for all points before the actual simplification process. It performs preliminary operations such as calculating normal vectors, curvature values, and density estimates for each point, storing these as metadata. This preliminary preparation enables the subsequent simplification step to make rapid decisions about which points to retain without performing complex real-time computations during the reduction process.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent applies self-service by using the point cloud's own geometric properties to guide its simplification. The significance metric is computed intrinsically from the point cloud data itself (local density, curvature, distance to neighbors) without requiring external reference data or complex external algorithms. The point cloud essentially evaluates and simplifies itself based on its inherent geometric characteristics, reducing the need for external computational resources.

Inventive Principle:
Principle #25Self-service

Data Source

PatentEP3535662B1System and method for processing input point cloud having points
Publication Date: 2023.03.22 MITSUBISHI ELECTRIC CORP
  • EP3535662B1 patent drawingFigure 1
  • EP3535662B1 patent drawingFigure 2
  • EP3535662B1 patent drawingFigure 3A~3B

AI summary

Systems and methods for determining a pattern in time series data representing an operation of a machine. A memory to store and provide a set of training data examples generated by a sensor of the machine, wherein each training data example represents an operation of the machine for a period of time ending with a failure of the machine. A processor configured to iteratively partition each training data example into a normal region and an abnormal region, determine a predictive pattern absent from the normal regions and present in each abnormal region only once, and determine a length of the abnormal region. Outputting the predictive pattern via an output interface in communication with the processor or storing the predictive pattern in memory, wherein the predictive pattern is a predictive estimate of an impending failure and assists in management of the machine.