Camera Scene Reconstruction for Static and Dynamic Segmentation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Scene reconstruction using lidar is expensive, leading to high implementation costs.

Innovation Solution

Implement a scene reconstruction method and apparatus that utilizes a camera-based approach, including background segmentation, point-cloud sparse and dense reconstruction, and dynamic-static segmentation to separately process static and dynamic scenes, reducing reliance on costly lidar.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If lidar is used for scene reconstruction, then measurement precision and reliability are improved, but implementation cost increases significantly

Engineering Contradiction:
Improvescene reconstruction precisionVSAvoidimplementation cost
Core Design Contradiction:
Measurement precisionVSEase of manufacture

Solution Approach 1:

The patent replaces expensive lidar with cheap cameras that can be mass-produced. Multiple standard cameras are used instead of a single costly lidar sensor, achieving comparable reconstruction precision through software algorithms rather than expensive hardware

Inventive Principle:
Principle #27Cheap short-living objects (Disposable)

Solution Approach 2:

The patent substitutes the mechanical laser ranging system (lidar) with an optical imaging system (camera) combined with computational algorithms. The reconstruction is achieved through image processing and geometric computation rather than direct mechanical measurement

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Ease of manufacture

If camera-based reconstruction is used, then implementation cost is reduced, but measurement precision and reliability deteriorate

Engineering Contradiction:
Improveimplementation costVSAvoidscene reconstruction precision
Core Design Contradiction:
Ease of manufactureVSMeasurement precision

Solution Approach 1:

The patent divides the scene into static and dynamic segments, processing each with appropriate algorithms. Static scenes use structure-from-motion (SfM) while dynamic scenes use optical flow methods, improving overall precision by tailoring processing to scene characteristics

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent combines multiple camera views and multiple processing algorithms (SfM, optical flow, depth estimation) to compensate for individual limitations. By merging information from multiple sources, the system achieves precision comparable to lidar

Inventive Principle:
Principle #5Merging (Combining)

3Measurement precision

If dynamic-static segmentation is performed, then reconstruction accuracy for moving objects is improved, but processing time and complexity increase

Engineering Contradiction:
Improvedynamic scene reconstruction accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent performs dynamic-static segmentation early in the processing pipeline, before detailed reconstruction. This preliminary classification allows subsequent processing to be optimized for each scene type, reducing overall processing time while maintaining accuracy

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP4708206A1Scene reconstruction method and apparatus, and storage medium and electronic device
Publication Date: 2026.03.11 BEIJING HORIZON INFORMATION TECH CO LTD
  • EP4708206A1 patent drawingFigure 1-1~1-2
  • EP4708206A1 patent drawingFigure 2~3
  • EP4708206A1 patent drawingFigure 4~5-1

AI summary

Disclosed in the embodiments of the present disclosure are a scene reconstruction method and apparatus, a storage medium, and an electronic device. The method includes: performing background segmentation on a first image sequence captured by a camera provided on a mobile device, to obtain a second image sequence; performing point-cloud sparse reconstruction by using the second image sequence, to obtain a first reconstructed point cloud and a first camera-pose sequence of the camera; performing dynamic-static segmentation on the first image sequence to obtain a third image sequence corresponding to a static scene and a fourth image sequence corresponding to a dynamic scene; performing point-cloud dense reconstruction by using the first reconstructed point cloud, the first camera-pose sequence, and the third image sequence, to obtain a second reconstructed point cloud; sequentially performing point-cloud sparse reconstruction and dense reconstruction by using the first camera-pose sequence and the fourth image sequence, to obtain a third reconstructed point cloud; and determining, based on the second reconstructed point cloud and the third reconstructed point cloud, a scene reconstruction result corresponding to the first image sequence.