Object Counter Using Area Flow Integration for Clustered Objects
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current object detection algorithms struggle to accurately count objects that are close together or moving in clusters, leading to underestimation, especially in environments with limited processing power, such as embedded systems.
Innovation Solution
A method that captures images of moving objects and calculates the integral of motion region areas to estimate object counts, using a contribution zone and area flow approach, which reduces processing intensity and bandwidth requirements, allowing for accurate counting without the need for object tracking.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If object detection algorithms are used to count objects, then objects can be detected and identified, but objects close together or moving in clusters are detected as single objects leading to underestimation
Solution Approach 1:
The patent segments the counting process into discrete temporal intervals using a time-discrete area flow model. Instead of attempting to segment spatially overlapping objects simultaneously, the system divides time into intervals and calculates the area flow crossing the boundary during each interval, effectively segmenting the measurement process temporally to avoid clustering issues
Solution Approach 2:
The patent replaces complex mechanical object tracking and identification systems with a simplified area flow calculation approach. By substituting the need for sophisticated object detection and tracking algorithms with a mathematical integration of area flow over time, the system achieves accurate counting without requiring advanced image processing capabilities
2Measurement precision
If shape analysis is used to estimate object count in clusters, then counting accuracy improves, but processing power requirements increase significantly
Solution Approach 1:
The patent extracts only the essential information needed for counting—the area flow crossing the boundary—while discarding all other complex object features. By taking out only the necessary measurement (area flow integration) and eliminating the need for shape analysis, object identification, and tracking, the system achieves accurate counting with minimal processing power
Solution Approach 2:
The patent changes the measurement parameter from object-based metrics (shape, size, position) to area flow-based metrics. By transforming the counting problem from detecting individual objects to integrating area flow over time, the system achieves equivalent or superior accuracy with dramatically reduced computational requirements
3Measurement precision
If intensive processing operations are performed to count clustered objects, then counting accuracy improves, but bandwidth requirements and system complexity increase
Solution Approach 1:
The patent performs preliminary integration of area flow during the counting process itself, rather than requiring post-processing operations. By accumulating area flow values in discrete time intervals as objects cross the boundary, the system prepares the count data in advance, eliminating the need for complex subsequent processing operations
Data Source
AI summary
An object counter performs a method for estimating the number of objects crossing a counting boundary. The method comprising: capturing, during a time period, a plurality of images representing moving images; registering, from the captured images, motion region areas passing across the counting boundary; calculating the integral of the registered motion region areas for forming a resulting total motion region area; and estimating the number of objects that have crossed the counting boundary by dividing the resulting total motion region area by a reference area.