Hybrid Ego-Motion Estimation for UAV Video Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for estimating camera ego-motion from videos captured by moving platforms, such as UAVs or helicopters, are highly computation-intensive and require feature detection and optical flow, making them inefficient for real-time processing in intelligence, reconnaissance, and surveillance applications.
Innovation Solution
A hybrid method combining a projected correlation method for initial ego-motion estimation and a valley search or alternate line search method for optimal estimation, which reduces computational cost while maintaining accuracy, and detects independent moving objects in the scene.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If feature detection and optical flow methods are used for ego-motion estimation, then measurement precision is improved, but productivity deteriorates due to high computational intensity
Solution Approach 1:
The patent segments the ego-motion estimation process into two distinct phases: (1) a fast initial estimation phase using simple image differencing and correlation methods to obtain a rough motion estimate, and (2) a refinement phase using optical flow only in a localized search region around the initial estimate. This segmentation allows the computationally intensive optical flow method to be applied only where necessary, significantly reducing overall computational cost while maintaining accuracy.
Solution Approach 2:
The patent performs preliminary action by computing a rough ego-motion estimate using computationally inexpensive image differencing and correlation methods before applying the more expensive optical flow method. This preliminary estimate serves as a starting point that guides subsequent refinement, allowing the system to avoid performing full optical flow computation over the entire image and instead focus only on a localized search region.
2Measurement precision
If feature detection and optical flow are used for ego-motion estimation, then measurement precision is improved, but device complexity worsens
Solution Approach 1:
The algorithm is segmented into distinct functional modules: image differencing, correlation-based initial estimation, and localized optical flow refinement. Each module has a specific, simple function, and they are combined in a pipeline fashion. This modular segmentation reduces overall algorithmic complexity compared to using full optical flow methods while maintaining measurement precision through the coordinated operation of these simpler components.
3Measurement precision
If conventional ego-motion estimation methods are used, then measurement precision is improved, but loss of time worsens due to computational intensity
Solution Approach 1:
The system performs preliminary action by quickly computing an initial ego-motion estimate using simple image differencing and correlation methods before applying the more time-consuming optical flow method. This preliminary estimate allows the system to skip time-intensive computations and directly focus refinement efforts on a small search region, dramatically reducing total processing time while maintaining accuracy.
Solution Approach 2:
The processing is segmented into a fast initial estimation phase using image differencing and correlation, followed by a targeted refinement phase. This time-efficient segmentation allows the system to obtain a sufficiently accurate estimate quickly, and only then invest additional time in refinement where truly necessary, minimizing overall loss of time.
Data Source
AI summary
Described is a system for compensating for ego-motion during video processing. The system generates an initial estimate of camera ego-motion of a moving camera for consecutive image frame pairs of a video of a scene using a projected correlation method, the camera configured to capture the video from a moving platform. An optimal estimation of camera ego-motion is generated using the initial estimate as an input to a valley search method or an alternate line search method. All independent moving objects are detected in the scene using the described hybrid method at superior performance compared to existing methods while saving computational cost.