Gas Cloud Spatial Tracking via Stereo Optical Flow
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods fail to accurately determine spatial information, including movement, of semi-transparent dynamic gaseous structures like gas clouds under real-world conditions without a known background, as they rely on background features for segmentation and depth information.
Innovation Solution
A method using a pair of cameras or a single camera with a beam splitter to record gaseous structures, reducing noise and interference, deriving characteristics through differential images and optical flow, and merging features to generate a disparity image for movement information, allowing gas clouds to be distinguished from backgrounds without prior background determination.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If stereo matching technique on difference images is used, then depth information can be obtained, but the method fails under uncontrolled conditions without known background
Solution Approach 1:
The patent extracts and utilizes intrinsic characteristic information from the gas cloud itself (through differential images and optical flow) rather than relying on background features. This extraction of self-characteristics enables the method to work independently of background conditions.
Solution Approach 2:
The patent introduces differential images and optical flow as intermediary representations that capture the dynamic characteristics of the gas cloud. These intermediaries serve as the basis for correspondence determination, replacing the need for direct background-based segmentation.
2Difficulty of detecting and measuring
If background features are used for segmentation, then gas cloud can be segmented, but the method requires previously known background
Solution Approach 1:
The patent performs preliminary processing of the image pair to generate differential images and optical flow fields before correspondence determination. This preliminary extraction of dynamic characteristics prepares the data for direct matching without requiring background determination.
Solution Approach 2:
The gas cloud's own dynamic characteristics (captured through differential images and optical flow) are used to enable its own segmentation and detection. The method is self-sufficient and does not require external background information or preliminary background determination steps.
3Adaptability or versatility
If intrinsic characteristic information is extracted and fused, then spatial information can be derived without background, but the processing complexity increases
Solution Approach 1:
The patent segments the image processing into distinct computational stages: generating differential images, calculating optical flow, extracting characteristic information, and performing correspondence determination. This segmentation of processing steps manages complexity through structured computation.
Solution Approach 2:
The patent transforms the problem from direct spatial matching to a multi-dimensional feature space by computing differential images and optical flow fields. This dimensionality transformation allows correspondence to be determined in feature space rather than direct image space, handling the complexity through mathematical transformation.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables accurate determination of gas cloud spatial information, including distance and movement, under uncontrolled conditions, suppressing background influence and allowing real-time tracking of gas clouds regardless of background homogeneity.
Implementation Method 1
a single camera with a beam splitter, via which the gaseous structure is recorded from the first recording axis and the second recording axis
Implementation Method 2
Pre-filtering of the recorded images, with at least one image noise and/or other interference effects being reduced from the recorded images
Implementation Method 3
Deriving characteristics of the gaseous structure from the recorded images by means of a) differential images comprising pixel-by-pixel changes in concentration over time in the gaseous structure
Implementation Method 4
b) images of the amount of the optical flow comprising spatial shifts in concentration in the gaseous structure
Implementation Method 5
Merging and determining the correspondences of the derived features by means of a correspondence determination and generating a disparity image and deriving movement information about the gaseous structure from the spatial information using the disparity image
Data Source
Figure 1
Figure 2
AI summary
The invention relates to a method for determining spatial information of a gaseous structure (10), in particular a gas cloud, wherein the method comprises at least the following steps: providing (50) a first camera (1) for recording the gaseous structure (10) from a first recording axis (A) and a second camera (2) for recording the gaseous structure (10) from a second recording axis (B), or providing a single camera with a beam splitter, via which the gaseous structure (10) is recorded from the first recording axis (A) and the second recording axis (B), capturing (100) a first recording image (A1) from the direction of the first recording axis (A) and a second recording image (B1) from the direction of the second recording axis (B), deriving features (102) about the gaseous structure (10) from the recording images (A1,B1) by means of a) difference images comprising concentration changes over time in the gaseous structure (10) and b) optical flow images comprising spatial concentration shifts in the gaseous structure (10), fusing and determining (103) the correspondences of the derived features by means of a correspondence determination and generating a disparity image and deriving motion information about the gaseous structure (10) from the spatial information (104) based on the disparity image.