Bird's-Eye View Data Generation for Dynamic Motion Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques fail to reconstruct the motion of observing cameras and peripheral moving bodies, are limited to simple motions, and struggle with unstable GPS reception due to high-rise structures, and cannot generate bird's-eye view data without static landmarks or when images from a bird's-eye viewpoint are not acquired.
Innovation Solution
A bird's-eye view data generating device that acquires time-series data from a dynamic environment, using a motion model and prior information about object sizes to generate data expressing the on-ground locus of movement of observing and peripheral bodies, even without static landmarks, by tracking and optimizing energy cost functions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional monocular 3D pedestrian localization techniques are used, then position estimation can be achieved, but the motion of the observing camera and loci of movement of peripheral moving bodies cannot be reconstructed
Solution Approach 1:
The patent combines multiple techniques (monocular 3D pedestrian localization, CubeSLAM, and GNSS) into a unified bird's-eye view data generation system that simultaneously reconstructs the motion of the observing camera and peripheral moving bodies, resolving the information loss in conventional single-technique approaches
2Reliability
If CubeSLAM technique is used, then self-position estimation can be achieved, but it is applicable only to environments with stable observation of moving bodies and static landmarks, and cannot handle complex interactions
Solution Approach 1:
The patent creates a multi-functional system that can operate in diverse environments (with or without static landmarks, with stable or unstable GPS) by integrating multiple estimation techniques and selecting appropriate methods based on environmental conditions, achieving both reliability and versatility
3Measurement precision
If GNSS technique is used, then self-position can be estimated, but positions of peripheral moving bodies cannot be reconstructed and GPS reception is unstable in high-rise environments
Solution Approach 1:
The patent uses visual observation data and motion models as intermediary information to supplement and correct GNSS positioning, enabling accurate position estimation of both the observing device and peripheral bodies even when GPS reception is unstable due to high-rise structures
4Measurement precision
If JP-A No. 2021-77287 technique is used, then bird's-eye view data can be generated, but it cannot be applied when images from bird's-eye viewpoint are not acquired
Solution Approach 1:
The patent creates a virtual bird's-eye view copy of the actual scene by processing first-person viewpoint images through 3D reconstruction and coordinate transformation, enabling bird's-eye view data generation without requiring actual bird's-eye cameras, thus expanding applicability
Data Source
AI summary
The present disclosure provides a bird's-eye view data generating device including a memory; and at least one processor coupled to the memory.