Crowd Image Target Detection Using Movement and Congestion Cues
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image analysis techniques fail to accurately detect a detection target when it is hidden behind another person in a crowd due to the absence of its distinctive features in the image.
Innovation Solution
An image processing apparatus that determines the degree of congestion, line of movement, and moving velocity of individuals in a series of images to estimate the presence of a detection target based on these factors, rather than relying solely on the target's appearance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image analysis relies on distinctive features of the detection target, then detection accuracy is improved when the target is visible, but detection capability deteriorates when the target is hidden behind another person
Solution Approach 1:
The patent introduces an intermediary approach by detecting the target indirectly through the movement patterns and spatial relationships of surrounding people. Instead of directly observing the hidden target, the system uses the movements of visible people as mediators to infer the presence and location of the hidden target, thereby resolving the detection capability issue when the target is obscured.
Solution Approach 2:
The patent inverts the traditional detection approach by not searching for the target's distinctive features directly, but rather searching for the absence of people in expected movement paths. This inversion allows the system to detect hidden targets by identifying areas where people should be moving but are not, thus solving the detection accuracy problem for hidden targets.
2Reliability
If the system analyzes movement patterns and congestion levels, then detection capability for hidden targets is improved, but processing complexity increases
Solution Approach 1:
The patent segments the image into multiple regions and analyzes movement patterns and congestion levels separately in each region. This segmentation approach allows the complex detection task to be divided into manageable sub-tasks, improving detection capability in crowded scenes while controlling processing complexity through localized analysis rather than global processing.
Solution Approach 2:
The patent applies partial action by focusing the analysis only on specific regions and movement patterns that are most indicative of hidden targets, rather than processing the entire image uniformly. This selective processing approach improves detection reliability for hidden targets while reducing overall processing complexity by avoiding unnecessary analysis of irrelevant areas.
Data Source
AI summary
To detect a detection target by an image analysis, the present invention provides an image processing apparatus 10 including: a determination unit 11 that determines at least one of a degree of congestion, a line of movement, and a moving velocity of a person included in a plurality of images being consecutive in time series; and a presence estimated part detection unit 12 that detects, from the image, a presence estimated part being a part where a detection target is estimated to be present, based on a result of the determination.