Safety Belt Detection Network for Accurate Wear and Position
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing safety belt detection methods in driver monitoring systems suffer from low accuracy, high computational load, and difficulty in determining proper wearing of safety belts due to environmental factors and the slender, changeable shape of safety belts.
Innovation Solution
A safety belt detection method using a deep learning-based detection network comprising an image classification branch network and an image segmentation branch network to quickly determine if a driver is wearing a safety belt and identify its position, employing semantic segmentation and reduced-size image processing to enhance accuracy and reduce computational load.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep learning-based image classification and segmentation networks are used for safety belt detection, then detection accuracy and position information are improved, but computational load and processing time increase
Solution Approach 1:
The detection network is divided into two separate branches: an image classification branch for determining whether the driver is wearing a safety belt, and an image segmentation branch for obtaining position information of the safety belt. This segmentation allows each branch to be optimized independently, improving overall detection accuracy while managing computational load through specialized processing paths.
2Measurement precision
If full-size images are processed for safety belt detection, then detection detail accuracy is improved, but computational overhead increases
Solution Approach 1:
The patent applies semantic segmentation only to specific regions of the image rather than processing the entire full-size image. The segmentation branch focuses computational resources on areas where the safety belt is likely to appear, obtaining sufficient position information while significantly reducing computational overhead compared to processing complete high-resolution images.
3Ease of manufacture
If existing detection methods are used for safety belt detection, then implementation simplicity is maintained, but detection accuracy and detail information are insufficient
Solution Approach 1:
The patent merges image classification and image segmentation techniques into a unified detection network that shares a common backbone. This combination allows the system to leverage the simplicity of classification while incorporating the detailed position information capability of segmentation, achieving high detection accuracy without completely redesigning the implementation from scratch.
Data Source
AI summary
A safety belt detection method, apparatus, computer device, and computer readable storage medium are disclosed. In the detection method, an image to be detected is obtained. The image to be detected is inputted into a detection network which includes an image classification branch network and an image segmentation branch network. A classification result, which indicates whether a driver is wearing a safety belt and is output from the image classification branch network, is obtained. A segmentation image, which indicates a position information of the safety belt and is output from the image segmentation branch network, is obtained. A detection result of the safety belt, indicating whether the driver wears the safety belt normatively, is obtained based on the classification result and the segmentation image.