Cloud Occlusion Prediction Using Dense Optical Flow Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing cloud occlusion prediction methods struggle with continuous tracking and feature point updating due to changing cloud shapes and light source interference, leading to inaccurate predictions of cloud occlusion times in solar thermal power generation systems.
Innovation Solution
A cloud occlusion prediction method using a dense optical flow method that tracks the entire cloud cluster, performs dense optical flow calculation, and corrects velocity using a sliding window algorithm to predict cloud occlusion times accurately.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If sparse optical flow method (Lucas-Kanade) is used to track cloud images, then calculation complexity is reduced, but tracking accuracy deteriorates due to inability to handle continuous cloud shape changes and feature point updates
Solution Approach 1:
The patent changes the fundamental parameter of optical flow density from sparse (feature points only) to dense (all pixels), enabling continuous tracking of cloud shape changes. This is achieved by applying dense optical flow algorithm to calculate motion vectors for all pixels in the image, not just extracted feature points, thereby maintaining tracking accuracy during continuous cloud deformation.
Solution Approach 2:
The patent segments the cloud tracking problem into two distinct phases: cloud detection/identification phase and cloud tracking phase. In the detection phase, cloud regions are identified and segmented from the background. In the tracking phase, dense optical flow is applied specifically to the detected cloud regions. This segmentation allows the system to handle both static and dynamic cloud shapes effectively.
2Power
If feature point method is used for cloud tracking, then calculation amount is reduced, but reliability deteriorates due to feature point changes and inability to track newly appeared cloud clusters
Solution Approach 1:
The patent ensures continuous tracking by applying dense optical flow to all pixels throughout the entire cloud region, not just discrete feature points. This continuous approach allows the system to maintain reliable tracking even when clouds deform, split, or new clouds appear, as every pixel contributes to the tracking information rather than relying on a limited set of feature points that may disappear or change.
3Ease of operation
If cloud clusters are treated as shape-constant objects for prediction, then prediction simplicity is improved, but prediction accuracy deteriorates due to continuous cloud shape changes
Solution Approach 1:
The patent transitions from treating clouds as static, shape-constant objects to dynamic objects with continuously changing shapes. Dense optical flow is applied to capture the motion of all pixels in cloud regions, allowing the system to track and predict cloud positions accurately even as they deform, move, and change shape over time. This dynamic approach replaces the simplistic shape-constant assumption with a flexible, adaptive tracking method.
Data Source
AI summary
The present invention discloses a cloud occlusion prediction method based on a dense optical flow method. The method disclosed by the present invention can better predict the cloud occlusion time and improve the tracking accuracy by tracking the movement of the entire cloud cluster, and can continuously track the newly-appeared cloud cluster.