Foreground Detection Under Illumination Variation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing methods fail to accurately detect foreground objects under varying environment illumination, especially when visual elements are occluded, leading to false detection and reduced accuracy in foreground detection.
Innovation Solution
The method updates both matching and non-matching visual element models in the scene model immediately following changes in environment illumination, using a first updating unit for matching models and a second updating unit for non-matching models, ensuring the scene model adapts to illumination variations, thereby preventing false detection of occluded elements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional background subtraction method is used to detect foreground, then detection speed is maintained, but false foreground detection occurs under illumination variation
Solution Approach 1:
The patent applies preliminary action by pre-establishing a scene model that includes multiple visual element models representing different appearance states (occluded, unoccluded, different illuminations). Before foreground detection, the system prepares these pre-trained models to quickly compare against current video frames, enabling rapid adaptation to illumination changes without complex real-time calculations.
Solution Approach 2:
The patent employs parameter changes by dynamically adjusting the scene model parameters based on detected visual elements. When a visual element is detected, the system updates the appearance features (color, texture, illumination parameters) of corresponding visual element models in the scene model, allowing the model to adapt to changing illumination conditions and reduce false foreground detection.
2Adaptability or versatility
If visual element model is updated only when matching current frame, then model stability is maintained, but adaptation to illumination changes is delayed
Solution Approach 1:
The patent applies segmentation by dividing the scene model into multiple independent visual element models, each representing different appearance states (occluded, unoccluded, different illuminations). This segmentation allows selective updating of specific visual element models based on detection results, maintaining overall model stability while enabling targeted adaptation to illumination changes.
Solution Approach 2:
The patent implements dynamics by making the scene model adaptive and dynamic rather than static. The system continuously updates visual element models in the scene model based on detected visual elements and their appearance features, allowing the model to dynamically adapt to changing illumination conditions while maintaining stability through structured update mechanisms.
3Measurement precision
If strict threshold comparison is used for foreground detection, then detection precision is maintained, but false foreground occurs when element reappears after occlusion
Solution Approach 1:
The patent applies copying by creating and maintaining multiple copies of visual element models representing different appearance states (occluded, unoccluded, different illuminations) in the scene model. When a visual element reappears after occlusion, the system can compare against the appropriate model copy that matches the current appearance state, reducing false foreground detection while maintaining detection precision.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present invention discloses an image processing apparatus and method and storage medium storing instructions. The image processing apparatus comprises a unit for acquiring a current video frame and a scene model based on an input video; a unit for updating visual element models in said scene model, which match with the visual elements in said current video frame, based on the visual elements in said current video frame; and a unit for updating visual element models in said scene model, which do not match with the visual elements in said current video frame, based on the visual elements in said current video frame and in the previous video frames thereof. According to the present invention, the whole scene model can vary immediately following the variation in the environment illumination.