Human Detection Accuracy in Variable Crowd Density
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing human detection systems face accuracy issues in densely populated areas where human shapes are hidden by other objects, and in sparsely populated areas where training data includes crowd images, leading to reduced detection accuracy.
Innovation Solution
An image processing apparatus that combines first and second detection processing units, where the first unit uses a trained model for human detection and the second unit employs pattern matching, with a determination unit to correct detection results and a calculation unit to calculate the number of humans, allowing for improved accuracy by adapting to different population densities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If human detection uses a trained model relating to humans, then detection accuracy is improved in places with little overlapping of humans, but detection accuracy reduces in densely populated places where human shapes are hidden by other objects
Solution Approach 1:
The detection system is segmented into multiple specialized detection units: a first detection unit using a trained model for human detection, a second detection unit using pattern matching, and a determination unit. Each unit handles specific detection scenarios, allowing the system to adapt to different population densities by selecting or combining appropriate detection methods.
Solution Approach 2:
The detection system achieves multi-functionality by integrating multiple detection approaches within a single apparatus. The first detection unit handles sparse crowd scenarios with trained models, while the second detection unit handles dense crowd scenarios with pattern matching, making the system universally applicable across different population density conditions.
2Measurement precision
If human detection uses pattern matching, then detection accuracy is improved in densely populated places, but detection accuracy reduces in sparsely populated places because images including a crowd of objects are used as training data
Solution Approach 1:
Different detection methods are applied to different local conditions: the first detection unit using trained models is optimized for sparse crowd environments, while the second detection unit using pattern matching is optimized for dense crowd environments. The determination unit selects the appropriate detection method based on the local population density characteristics of the input image.
3Measurement precision
If multiple detection methods are integrated, then detection accuracy is improved across different population densities, but device complexity increases
Solution Approach 1:
The determination unit serves as an intermediary that coordinates between the first detection unit and the second detection unit. It receives input images, determines the appropriate detection method based on population density analysis, and directs the corresponding detection unit to process the image, thereby managing system complexity through centralized control.
Data Source
AI summary
An image processing apparatus includes a first detection unit that executes first detection processing for detecting an object presumed to be a human from an image by using a trained model relating to humans, a second detection unit that executes second detection processing for detecting an object presumed to be a human from the image at least by using pattern matching, a determination unit that determines whether the object detected through the first detection processing corresponds to the object detected through the second detection processing, a correction unit that corrects a detection result relating to the object that is detected through the first detection processing and is determined to correspond to the object detected through the second detection processing, and a calculation unit that calculates the number of humans in the image based at least the detection result corrected by the correction unit.