Image Processing Apparatus for Stationary Subject Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional subject detection methods, such as the background difference method, face challenges in accurately and efficiently detecting subjects, especially when the subject remains stationary for a period, leading to false detection of background changes and widened search ranges, which hampers rapid detection.
Innovation Solution
An image processing apparatus that calculates differences between input video data and a background model, updates the model based on appearance information, and selectively updates the standard model to maintain accurate background representation, thereby reducing false positives and narrowing the detection search range.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the background difference method is used to detect moving objects, then the detection can be performed based on feature quantity differences, but when a subject remains stationary for a predetermined time, the background features are forgotten from the standard model, causing false detection of background changes and widened search ranges
Solution Approach 1:
The patent applies local quality by differentiating between foreground and background regions based on spatial location and temporal characteristics. The determination device analyzes local feature quantities in specific regions (detection windows) and compares them with standard models for both foreground and background, allowing accurate identification of stationary subjects without false detection of background changes.
Solution Approach 2:
The patent implements dynamics by making the standard model adaptive and updateable. The update device dynamically adjusts the standard model based on detected subject movements and background changes, allowing the system to respond to temporal variations in the scene while maintaining accurate detection of stationary objects.
2Reliability
If subject detection is performed by scanning detection windows in the entire image, then comprehensive coverage is achieved, but processing costs become very expensive
Solution Approach 1:
The patent applies segmentation by dividing the image into multiple detection windows and systematically processing only the relevant portions. The determination device scans detection windows in a structured manner, and the update device selectively updates only the standard model regions affected by detected subjects, reducing overall processing requirements while maintaining detection completeness.
Solution Approach 2:
The patent implements partial action by performing subject detection only in regions where background differences indicate potential subjects. The determination device identifies candidate regions based on feature quantity comparisons, and subsequent processing is concentrated only on these partial regions rather than the entire image, significantly reducing processing costs.
3Productivity
If the search range for subject detection is narrowed based on background difference, then processing efficiency improves, but when a subject is stationary, the background difference causes the search range to be inappropriately widened
Solution Approach 1:
The patent applies feedback by continuously monitoring detected subject positions and using this information to refine the standard model. The update device receives feedback from the detection device about stationary subjects and adjusts the standard model accordingly, preventing future false detection of background changes and ensuring accurate search range narrowing.
Solution Approach 2:
The patent implements preliminary action by pre-processing the image to identify potential subject regions before performing detailed subject detection. The determination device preliminarily compares feature quantities and identifies candidate areas, allowing the subsequent subject detection to focus only on these pre-identified regions with appropriate search range control.
Data Source
AI summary
An image processing apparatus includes a calculation unit configured to calculate a difference between a feature quantity of input video data and a feature quantity of a model representing a background, a determination unit configured to determine whether a partial area in the input video data is a foreground or a background based on the difference, a detection unit configured to detect a subject area from an area determined to be a foreground, a first update unit configured to update appearance information that represents an appearance state of a background relating to the subject area, and a second update unit configured to update the model based on the appearance information.