Person Detection Using HOG Feature Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In person detection systems mounted on moving bodies like forklift trucks, the accuracy of person detection decreases when only the upper body of a person is visible due to obstructions within the camera's imaging range, as existing methods struggle to differentiate between the person and other obstacles.
Innovation Solution
An image processing device that performs an upper body detection process by comparing areas with upper body comparison data and a whole-body detection process by comparing areas with whole-body comparison data, determining whether an obstacle is a person based on its distance from the road surface and features extracted using HOG features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a person detection process is performed using HOG features on image data that does not include the whole body of a person, then the detection process can still be executed, but the accuracy of person detection decreases
Solution Approach 1:
The detection process is segmented into two distinct modes: upper body detection mode and whole-body detection mode. The system determines which mode to use by checking whether the detected obstacle is within a predetermined range from the camera and is distanced from the road surface within a predetermined range. This segmentation allows the system to maintain high accuracy by selecting the appropriate detection mode based on the specific detection scenario.
Solution Approach 2:
The system applies different detection characteristics to different regions or conditions. When the obstacle meets the upper body detection process condition (within predetermined range from camera and distanced from road surface), the system uses upper body comparison data for detection. When the obstacle does not meet this condition, the system uses whole-body comparison data. This local quality approach ensures optimal detection accuracy for each specific condition.
2Loss of information
If the lower body of a person is hidden by an obstruction within the camera's imaging range, then the camera can still capture the upper body, but the system cannot accurately differentiate between the person and other obstacles
Solution Approach 1:
The system introduces an intermediary judgment mechanism that uses the result of the upper body detection process as a mediator to determine the detection mode. By first performing the upper body detection process and checking whether the obstacle meets the predetermined conditions, the system selects the appropriate comparison data (upper body comparison data or whole-body comparison data) to accurately differentiate between persons and obstacles even when the lower body is hidden.
3Measurement precision
If the system uses upper body comparison data for detection, then detection accuracy is maintained when the lower body is hidden, but the system complexity increases due to multiple detection processes
Solution Approach 1:
The detection system is made dynamic by automatically switching between upper body detection mode and whole-body detection mode based on real-time conditions. The system determines which mode to use by checking whether the obstacle is within a predetermined range from the camera and is distanced from the road surface within a predetermined range. This dynamic adaptation allows the system to maintain high accuracy without permanently implementing complex dual-mode architecture.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enhances the accuracy of person detection by comparing features at the same positions, reducing errors and maintaining detection precision even when the lower body is hidden by obstructions, without requiring high-performance hardware or dedicated deep learning for upper body detection.
Implementation Method 1
The person detection process is performed using HOG (Histogram of Oriented Gradients) features, for example
Data Source
AI summary
An image processing device of a person detection system mounted on a moving body is configured to: detect, in image data obtained from a camera, an area in which an obstacle appears; determine whether the area meets an upper body detection process condition that the obstacle in the area is distanced from a road surface within a predetermined range from the camera; perform an upper body detection process in which the area of the image data is compared with upper body comparison data to determine whether the obstacle in the area is a person, for the area that meets the upper body detection process condition; and perform a whole-body detection process in which the area of the image data is compared with whole-body comparison data to determine whether the obstacle in the area is a person, for the area that does not meet the upper body detection process condition.