Object Counting via Overlapping Estimation Windows
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object counting devices face challenges in accurately counting persons in images, particularly when individuals overlap, due to incorrect arrangement of estimation windows, leading to erroneous undercounting or double counting.
Innovation Solution
An object counting system that acquires and sets estimation areas in images to include overlapping objects, estimating the number of objects in each area and computing object density where areas overlap, ensuring accurate counting by optimally arranging estimation windows.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If estimation windows are arranged without overlapping, then device complexity is reduced and processing is simplified, but measurement precision deteriorates because persons smaller than the window size cannot be detected
Solution Approach 1:
The image is divided into multiple estimation windows that may overlap, allowing different regions to be processed independently while collectively covering the entire image space. This segmentation enables small persons to be detected in at least one window while maintaining manageable processing complexity through systematic arrangement.
Solution Approach 2:
The patent introduces a new dimension to the arrangement strategy by considering overlap as a valid and useful property rather than avoiding it. By transitioning from non-overlapping to overlapping arrangements, the system gains the ability to detect objects of varying sizes across different window configurations, solving the detection accuracy problem.
2Measurement precision
If estimation windows are arranged to overlap, then measurement precision improves by detecting smaller persons, but device complexity increases and double counting occurs
Solution Approach 1:
The patent implements feedback mechanisms where detection results from overlapping windows are compared and integrated. By analyzing detection outcomes across multiple windows and applying integration rules, the system resolves double counting issues while maintaining the benefits of overlapping arrangements for detecting smaller objects.
Solution Approach 2:
The system dynamically adjusts arrangement parameters such as window size, overlap degree, and detection thresholds based on image characteristics and detection results. This parameter optimization enables effective use of overlapping windows while minimizing complexity and preventing double counting through adaptive control.
3Device complexity
If a single fixed-size estimation window is used, then device complexity is minimized, but measurement precision deteriorates because it cannot adapt to persons of different sizes
Solution Approach 1:
The patent transitions from static fixed-size windows to dynamic arrangements where window parameters such as size and position can vary. This dynamic approach allows the system to adapt to different object sizes and distributions in the image, improving detection accuracy while maintaining reasonable processing complexity through systematic dynamic adjustment.
Data Source
AI summary
An object counting system includes an acquisition means for acquiring information of an estimation area which is a partial area of an image with which partial area a predetermined condition related to objects to be counted shown in the image are associated, the estimation area being a unit of area for estimating the number of the objects to be counted, a setting means for setting the estimation area in the image in such a manner that the estimation area indicated by the acquired information of the estimation area includes the objects to be counted which are not included in the objects to be counted in a different estimation area, and which satisfy the predetermined condition, an estimation means for estimating, in each estimation area, the number of the objects to be counted shown in the estimation area set in the image, and a computation means for computing a density of the objects to be counted in an area where predetermined areas in the estimation area are overlapped using the number of the objects to be counted that has been estimated in each estimation area.