Monocular Thermal Camera Motion Estimation in Low Contrast
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current computer vision systems fail to provide robust camera motion estimation and tracking in high-motion, low-contrast environments, particularly for first responders in hazardous situations where external tracking systems like GPS and cellular towers are unreliable or unavailable.
Innovation Solution
A monocular thermal camera system combined with an inertial measurement unit (IMU) estimates camera motion using visual tracking and sensor data, employing a hierarchical homography fitting approach to initialize and update 3D structure, even in low-contrast conditions, and integrates with wireless signal data for location determination.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Weight of moving object
If a monocular thermal camera system is used for first responder tracking, then the system weight is reduced and visibility is improved, but camera motion estimation accuracy deteriorates due to high motion and low contrast
Solution Approach 1:
The patent combines a monocular thermal camera with an inertial measurement unit (IMU) to create a fused sensor system. The IMU provides motion data that compensates for the thermal camera's inability to provide reliable feature matching in low-contrast environments. This merging allows the system to maintain accurate camera motion estimation despite the limitations of the lightweight monocular thermal camera.
Solution Approach 2:
The patent introduces an intermediary processing layer that fuses data from the thermal camera and IMU. The system uses the IMU's high-frequency motion data as an intermediary to bridge the gap between thermal image frames, enabling accurate motion estimation even when the thermal camera itself cannot provide sufficient feature correspondence due to low contrast.
2Measurement precision
If feature-based image tracking algorithms are used, then tracking accuracy improves in high-contrast environments, but performance deteriorates in low-contrast thermal images
Solution Approach 1:
The patent changes the fundamental parameters used for tracking by switching from feature-based methods to direct intensity-based comparison methods optimized for thermal images. The system uses algorithms specifically tuned for thermal imagery that do not rely on edge detection or feature matching, thereby maintaining tracking accuracy in low-contrast thermal environments while preserving adaptability across different thermal conditions.
3Measurement precision
If external tracking systems (GPS, cellular towers) are used, then location tracking accuracy is sufficient in urban environments, but reliability deteriorates in remote locations or disaster scenes
Solution Approach 1:
The patent implements a self-service tracking system that does not depend on external infrastructure. By using the monocular thermal camera combined with IMU data and employing algorithms that can operate independently without GPS or cellular tower support, the system provides autonomous location tracking and route monitoring that remains reliable in remote locations and disaster scenes where external systems are unavailable.
Data Source
AI summary
Systems and methods are provided that track camera motion from image and sensor data in single-camera, low contrast and high-motion systems. Camera motion is estimated through dense visual tracking using image and sensor data. Motion sensor data from a wearable motion sensor worn by a human is used to determine initial camera motion parameters. Image data from a thermal imaging camera outputting low contrast video frames is used for motion tracking. The camera motion in the frame is represented by a translation and a rotation of the camera through an environment. The frames are down-sampled to generate image pyramid of frames of progressively lower resolution. A hierarchical homography optimizing approach is described. A homography is optimized across each resolution level beginning with the lowest resolution frames. A modified translation and rotation displacement of the camera is determined based on the optimized homography.