Single-Camera Object Pose Estimation from Sensor Motion

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing vehicle sensor systems struggle to accurately determine the pose of stationary objects with minimal data, particularly when using a single sensor like a camera that does not provide depth data.

Innovation Solution

A computer system utilizes motion data from vehicle sensors to calculate virtual pixel locations from a virtual camera's perspective, enabling the determination of an object's pose with six degrees of freedom using a single camera, without requiring detailed object features or geometry data.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If a single sensor (e.g., camera) is used to detect objects, then device complexity is reduced, but measurement precision of object pose deteriorates due to lack of depth data

Engineering Contradiction:
Improvesensor system complexityVSAvoidobject pose measurement precision
Core Design Contradiction:
Device complexityVSMeasurement precision

Solution Approach 1:

The patent inverts the traditional approach by not moving the object into the sensor frame, but instead moving the sensor into the object frame. This allows a single camera to determine 6-DOF pose by treating the sensor motion as virtual camera movement, eliminating the need for multiple sensors or active object motion while maintaining measurement precision

Inventive Principle:
Principle #13The other way round (Inversion)

Solution Approach 2:

The patent introduces motion data as an intermediary element that bridges the gap between single-sensor limitations and 6-DOF pose requirements. By incorporating sensor motion information (from IMU, wheel encoders, or other motion sensors) as a mediator, the system compensates for the lack of depth data from a single camera and achieves accurate pose estimation

Inventive Principle:
Principle #24Intermediary (Mediator)

2Ease of operation

If minimal object data (single point tracking) is used, then ease of operation is improved, but information completeness deteriorates

Engineering Contradiction:
Improvedata processing simplicityVSAvoidobject feature information
Core Design Contradiction:
Ease of operationVSLoss of information

Solution Approach 1:

The patent extracts only the essential information needed for pose determination - a single tracked point on the object - while discarding unnecessary object features and geometry data. This extraction approach maintains ease of operation by simplifying data processing while the sensor motion compensation ensures complete pose information (6-DOF) is still obtained

Inventive Principle:
Principle #2Taking out (Extraction)

3Measurement precision

If sensor motion is utilized to compensate for limited sensor data, then measurement precision is improved, but device complexity worsens due to integration of motion sensors

Engineering Contradiction:
Improvepose determination accuracyVSAvoidsensor integration complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent merges the functionality of the camera with motion sensor data (IMU, wheel encoders, or other motion sensors) to create a unified pose estimation system. By combining visual data from the camera with motion compensation data, the system achieves 6-DOF pose determination using a single camera plus existing motion sensors, avoiding the need for complex multi-camera setups while maintaining measurement precision

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentUS12573075B2Object pose from sensor motion
Publication Date: 2026.03.10 FORD GLOBAL TECH LLC
  • US12573075B2 patent drawing
  • US12573075B2 patent drawing
  • US12573075B2 patent drawing

AI summary

A computer includes a processor and a memory, and the memory stores instructions executable to receive sensor data from a sensor indicating an object at a plurality of time steps, receive motion data indicating motion of the sensor at the time steps, determine a plurality of virtual pixel locations of the sensor at the time steps from a perspective of a virtual camera located at the object based on the sensor data and on the motion data, and determine a pose of the object using the virtual pixel locations. The sensor moves while generating the sensor data, and the object is stationary.