Person Counting via Shielding Object Association
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image analysis techniques for estimating the number of people in images struggle with accuracy when detection targets are partly or entirely shielded by objects, such as umbrellas, leading to incorrect counting during high congestion scenarios.
Innovation Solution
An image processing apparatus is designed with a first detection unit for identifying people, a second detection unit for identifying shielding objects, a specification unit for associating people with their shielding objects, and a determination unit to accurately count the number of people by dividing the image into small areas and using learned models for detection and association processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If simple detection methods are used to count people in images, then the device complexity is low, but the measurement precision deteriorates when detection targets are shielded by objects
Solution Approach 1:
The system segments the detection task into multiple independent detection units, each trained to detect specific object categories (e.g., persons, umbrellas, bags). This segmentation allows the system to handle shielded targets by detecting both the target and the shielding object separately, then reasoning about their relationships to improve counting accuracy without requiring a single complex detection model.
Solution Approach 2:
The system introduces an intermediary reasoning mechanism that connects detection results with counting outcomes. By detecting shielding objects and analyzing their spatial relationships with potential targets, the system mediates between simple detection and accurate counting, resolving the contradiction between device simplicity and measurement precision.
2Measurement precision
If detection models are trained to handle shielded targets, then the measurement precision improves, but the difficulty of detecting and measuring increases
Solution Approach 1:
Instead of training a single complex model to handle all shielding scenarios, the system segments the detection into multiple specialized models, each trained on specific object categories. This reduces the training difficulty for each individual model while improving overall detection accuracy through the combination of multiple specialized detectors.
Solution Approach 2:
The system changes the parameters of detection by using multiple detection models with different training parameters (object categories, detection thresholds) rather than a single model. This allows the system to adapt to different detection scenarios and improve accuracy without requiring each individual model to be excessively complex.
3Measurement precision
If multiple detection units are used to detect both people and shielding objects, then the measurement precision improves, but the device complexity increases
Solution Approach 1:
The system divides the detection function into multiple specialized detection units, each responsible for detecting specific object types. This segmentation improves measurement precision by allowing each unit to specialize in its target category, while the modular structure manages complexity through clear separation of functions.
Solution Approach 2:
The detection units are designed with multi-functionality, where each unit can detect its specific object type and also contribute to the overall counting decision. The system universally applies the same multi-unit detection architecture to handle various shielding scenarios, managing complexity through standardized multi-functional components.
Data Source
AI summary
An image processing apparatus includes a first detection unit configured to detect a person in an input image, a second detection unit configured to detect a specific object in the input image, a specification unit configured to specify a combination of the detected person and the detected object indicating an identical person, and a determination unit configured to determine a number of person in the input image, based on the detected person, the detected object, and the specified combination.