Movement State Estimation Using Regional Count Dynamics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In crowded environments, it is challenging to accurately detect and track individual persons and estimate the number of moving persons due to large overlaps in photographed images, low frame rates, and difficulties in recognizing individual passersby.
Innovation Solution
A movement state estimation device that uses temporally sequential images to estimate the quantity of monitoring targets in local regions and then estimates the movement state from the time-series change in these quantities, employing techniques such as particle filters and optical flow calculation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If individual person detection and tracking methods are used in crowded environments, then measurement precision can be maintained in open spaces, but detection reliability deteriorates due to large overlaps and low frame rates
Solution Approach 1:
The patent segments the detection task from the tracking task. It uses segmentation to divide the image into multiple regions and counts persons in each region independently, rather than attempting to track individual persons through crowded areas where overlaps make identification difficult.
Solution Approach 2:
The patent extracts only the necessary information for the specific application goal. Instead of extracting and tracking individual person identities and trajectories, it extracts only the count of persons in each region, which is sufficient for measuring person flow and crowd density without requiring reliable individual tracking.
2Reliability
If crowd-patch learning methods are used to recognize crowds, then detection reliability improves in crowded environments, but measurement precision deteriorates because individual movement states cannot be determined
Solution Approach 1:
The patent introduces temporal dynamics by analyzing changes in person counts across multiple frames. By comparing the number of persons in each region at different time points, it can infer movement states (entering, exiting, staying) without needing to track individual persons, thus maintaining both reliability in crowded environments and precision in movement estimation.
Solution Approach 2:
The patent adds the time dimension to the analysis. Instead of relying solely on spatial information from single frames or static crowd patches, it incorporates temporal information by measuring changes in person counts across multiple frames, enabling movement state estimation through the fourth dimension of time.
3Measurement precision
If individual person tracking is attempted in crowded environments, then measurement precision can be maintained for moving persons, but device complexity increases due to the need for sophisticated tracking algorithms
Solution Approach 1:
The patent extracts only the count information needed for movement state estimation, discarding the complex task of individual person identification and trajectory tracking. This simplification reduces device complexity while maintaining the ability to measure person flow and movement patterns at the regional level.
Solution Approach 2:
Instead of tracking individual persons, the patent uses regional count copies across multiple frames to infer movement. By comparing count values in the same region at different times, it achieves movement detection without the computational complexity of individual tracking algorithms.
Data Source
AI summary
[Problem] To provide a motion condition estimation device, a motion condition estimation method and a motion condition estimation program capable of accurately estimating the motion condition of monitored subjects even in a crowded environment. [Solution] A motion condition estimation device according to the present invention is provided with a quantity estimating means and a motion condition estimating means. The quantity estimating means uses a plurality of chronologically consecutive images to estimate a quantity of monitored subjects for each local region in each image. The motion condition estimating means estimates the motion condition of the monitored subjects from chronological changes in the quantities estimated in each local region.