Camera-Based Seat Belt Detection for Proper Occupant Restraint
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing seat belt detection systems are prone to inaccuracies due to the use of weight sensors and buckle sensors, which can be deceived by tools like fake buckles or improper seat belt usage, leading to ineffective encouragement of seat belt wearing.
Innovation Solution
A seat belt wearing determination method and apparatus that utilizes a camera to capture vehicle interior images, determining bounding boxes for the occupant's upper body and seat belt, calculating the overlapping degree to accurately assess seat belt usage through machine learning and deep learning algorithms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If weight sensors and buckle sensors are used to detect seat belt wearing status, then the detection system is simple and low-cost, but the detection accuracy is poor and can be deceived by fake buckles or improper usage
Solution Approach 1:
The patent replaces mechanical sensors (weight sensors and buckle sensors) with an optical detection system using cameras and image processing algorithms. The camera captures images of the occupant and seat belt, and machine learning algorithms analyze the images to determine proper seat belt wearing, eliminating the vulnerability to fake buckles and improper usage that plagues mechanical sensor systems.
Solution Approach 2:
The patent introduces an intermediary processing layer between the sensor and the detection logic - specifically, image processing and machine learning algorithms that analyze the visual appearance of the seat belt and occupant. This intermediary layer enables the system to understand the context and proper configuration of seat belt wearing, rather than relying on simple mechanical signals that can be fooled.
2Measurement precision
If buckle sensors are used to determine seat belt fastening status, then the system is easy to implement, but it cannot distinguish between proper and improper seat belt usage
Solution Approach 1:
The patent transitions from one-dimensional mechanical sensor signals to two-dimensional image data, providing rich visual information about the seat belt and occupant configuration. By analyzing the spatial relationship, orientation, and appearance of the seat belt across the image plane, the system can determine proper usage patterns that are invisible to simple buckle sensors.
Solution Approach 2:
The patent changes the detection parameters from simple binary sensor signals (fastened/not fastened) to multiple visual parameters including seat belt position, orientation, tension appearance, and relationship to body contours. These parameter changes enable the system to evaluate whether the seat belt is properly configured, not just whether it is fastened.
Data Source
AI summary
Introduced is a seat belt wearing determination apparatus including a camera configured to capture an interior of a vehicle, a controller configured to determine a first bounding box and a second bounding box from an image including a seated occupant captured by the camera and determine whether the seated occupant is wearing a seat belt based on the overlapping degree of the first and second bounding boxes, wherein the first bounding box corresponds to the upper body region of the seated occupant, and the second bounding box includes the seat belt detected from the captured image by the controller.