Stagnant Object Detection via Partial Region Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image analysis techniques struggle to accurately detect stagnant objects, particularly when their shape or size changes due to inflow or outflow of individuals, or when there is switching of persons or objects within the object, leading to degraded detection accuracy.
Innovation Solution
An image processing apparatus and method that acquire partial region information for each frame image, extracting partial regions based on conditions of continued presence of target objects at a predetermined level and uniformity of their aggregation state, allowing for high-accuracy detection of stagnant objects by focusing on each partial region rather than the entire detected object.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If detection is based on position or size of the entire detected object, then detection process is simple, but detection accuracy degrades when shape or size changes
Solution Approach 1:
The patent divides the detected object into multiple partial regions and evaluates each partial region independently for stagnation characteristics. This segmentation allows the system to maintain simple detection processes while achieving high accuracy even when the overall shape or size of the stagnant object changes, because local regions maintain their stagnation properties regardless of global changes.
2Ease of operation
If motion detection is based on characteristic points and motion vectors, then motion detection is straightforward, but detection accuracy degrades when persons or objects switch positions
Solution Approach 1:
The patent applies segmentation by evaluating motion characteristics in each partial region separately rather than tracking characteristic points across the entire object. This approach maintains ease of operation while improving accuracy during switching events, as each partial region's stagnation is determined by local motion patterns rather than global point tracking.
Solution Approach 2:
The patent implements local quality by assigning different evaluation criteria to different partial regions, where each region is assessed based on its own motion characteristics. This allows the system to accurately identify stagnant regions even when other parts of the object are moving or switching positions, as each region is evaluated on its local motion properties.
3Productivity
If detection considers the entire detected object as a unit, then processing is efficient, but detection accuracy degrades during inflow or outflow of persons
Solution Approach 1:
The patent segments the detected object into multiple partial regions that can be processed independently. This segmentation maintains processing efficiency through parallel evaluation of multiple regions while achieving high detection accuracy during inflow or outflow events, as each region's stagnation status is determined independently of changes in other regions.
Data Source
AI summary
To detect a stagnant object by an image analysis at high accuracy, the present invention provides an image processing apparatus 10 including: an acquisition unit 11 that acquires partial region information for each frame image, the partial region information indicating a situation of a target object in each of a plurality of partial regions within one image for each frame image; and an extraction unit 12 that extracts the partial region from a plurality of the partial regions, based on the partial region information for the each frame image, the partial region to be extracted satisfying at least one of a condition that presence of a plurality of the target objects continues at a predetermined level or higher and a condition that uniformity of a state of an aggregation constituted by a plurality of the target objects being present continues at a predetermined level or higher.