Image Processing Apparatus Background Update Timing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for extracting a foreground region from images fail to accurately separate foreground objects from backgrounds when there are dynamic changes in the background or ambient lighting conditions, such as sunlight or illumination, leading to difficulties in updating background images when objects, like people, are present for extended periods.
Innovation Solution
A system that groups multiple cameras based on their targets and schedules background updates only when the image capturing targets do not include human figures, using a virtual viewpoint image generation system to combine foreground and background images, and employs a background partial update function to selectively update regions based on user-defined settings.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If background image is updated at a set timing using the technique of Japanese Patent Laid-Open No. 2021-163303, then the background region can be correctly written as the background image and not extracted as foreground region, but the background image cannot be updated when objects are present for extended periods
Solution Approach 1:
The patent implements dynamic background update timing by switching between two modes: when no human figures are detected, the background image is updated at set timing to adapt to background changes; when human figures are detected, the background image is not updated to avoid capturing foreground objects as background. This dynamic adjustment resolves the contradiction between maintaining extraction accuracy and enabling background updates.
Solution Approach 2:
The system uses detection results of human figures as feedback to control the background image update process. The detection unit continuously monitors for human figures, and based on this feedback, the control unit decides whether to proceed with background image updates. This feedback mechanism ensures that background updates only occur when appropriate, maintaining both accuracy and adaptability.
2Reliability
If background image is not updated when human figures are present, then foreground objects are not incorrectly classified as background, but the background image becomes outdated when background changes occur
Solution Approach 1:
The system dynamically adjusts background image update timing based on the presence of human figures. When no human figures are detected, the background image is updated at set timing to maintain currentness and adapt to background changes. When human figures are detected, updates are suspended to prevent misclassification. This dynamic approach resolves the contradiction between maintaining accuracy and keeping the background image current.
Solution Approach 2:
The background image update operates periodically at set timing intervals, but this periodic action is conditionally activated based on detection results. When the detection unit confirms no human figures are present, the periodic update proceeds; otherwise, it is skipped. This conditional periodic action allows the system to balance between keeping the background current and maintaining extraction accuracy.
3Adaptability or versatility
If background image is updated frequently to adapt to dynamic background changes, then the background image remains current, but foreground objects may be incorrectly classified as background
Solution Approach 1:
The system takes preliminary anti-action by preventing background image updates when human figures are detected, before the misclassification can occur. The detection unit continuously monitors for human figures, and the control unit proactively suspends updates when figures are present, preventing the harmful effect of incorrectly classifying foreground objects as background. This resolves the contradiction by anticipating and preventing the problem before it occurs.
Solution Approach 2:
The system uses real-time feedback from the detection unit to control the frequency of background image updates. When human figures are detected, the feedback signal prevents further updates, ensuring accuracy is maintained. When no figures are present, updates proceed at set timing to adapt to background changes. This feedback-controlled update mechanism resolves the contradiction between adaptability and accuracy.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An image processing apparatus (21) obtains a plurality of captured images obtained by image capturing of a plurality of image capturing units, extracts foreground regions from the plurality of captured images by a background difference method, and generates background images used in the background difference method while updating the background images based on the plurality of captured images. In the generation, the background images are updated according to an update method of the background images set for each of groups to which the plurality of image capturing units are classified based on targets of interest of the respective image capturing units in the image capturing, by using the captured images obtained by one or a plurality of the image capturing units included in the group.