Plane Detection Using Weighted RANSAC and Image Likelihoods

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing plane detection techniques, such as those described in PTL 1, lack the accuracy required for detecting planes in targets with complex geometries or occlusions, leading to suboptimal performance in identifying planes with high precision.

Innovation Solution

A plane detecting device and method that combines visible image information and 3D coordinate data using a robust estimation method, where likelihoods of planarity are calculated using a machine learning model to enhance the detection accuracy of planes, particularly by employing RANSAC with likelihoods as weights for improved estimation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If a conventional plane detection technique (e.g., RANSAC without likelihoods) is used, then the detection process is simple and fast, but the detection accuracy is insufficient for complex geometries or occlusions

Engineering Contradiction:

Improveplane detection accuracyVSAvoiddetection process complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent changes the parameter space by introducing likelihood values that represent the probability of each pixel belonging to the target plane. This transformation from binary classification to probabilistic assessment enables more nuanced and accurate plane detection, especially in complex scenarios with occlusions or varying geometries

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent performs preliminary classification of pixels into in-plane and out-of-plane categories using a machine learning model before applying the RANSAC algorithm. This preliminary action provides weighted likelihoods that guide the subsequent plane fitting process, improving accuracy by pre-identifying relevant pixels

Inventive Principle:

Principle #10Preliminary action

2Measurement precision

If visible image information and 3D coordinate information are combined with likelihoods, then plane detection accuracy improves, but computational complexity increases

Engineering Contradiction:

Improveplane detection accuracyVSAvoidcomputational power

Core Design Contradiction:

Measurement precisionVSPower

Solution Approach 1:

The patent applies partial action by using a machine learning model to classify only whether pixels are in-plane or out-of-plane, rather than performing full semantic segmentation. This partial classification provides sufficient likelihood information for plane detection while maintaining computational efficiency

Inventive Principle:

Principle #16Partial or excessive action

Solution Approach 2:

The patent merges multiple data sources (visible image information and 3D coordinate information) into a unified plane detection framework. By combining these modalities with likelihood weighting, the system achieves improved accuracy without requiring separate processing pipelines

Inventive Principle:

Principle #5Merging (Combining)

Data Source

PatentUS20240013421A1Plane detecting device, and plane detecting method

Publication Date: 2024.01.11 PANASONIC INTELLECTUAL PROPERTY MANAGEMENT CO LTD

AI summary

A plane detecting device according to the present disclosure includes an information acquisition unit, a likelihood acquisition unit, and a plane detector. The information acquisition unit acquires visible image information of a target having a predetermined plane and 3D coordinate information corresponding to the visible image information. The likelihood acquisition unit acquires likelihoods indicating a planarity of the predetermined plane of the target from the visible image information. The plane detector detects the predetermined plane of the target through a robust estimation method by using the 3D coordinate information and the likelihood.