In-Vehicle Pose Detection Using Bounding Areas for Seat Assignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current seat occupancy detection systems in vehicles face challenges in accurately assigning a person to a seat due to unclear positions or movements, leading to false assignments and potential safety issues, as they rely on weight sensors and 2D object detection which may not account for leaning postures.
Innovation Solution
A computerized method for pose detection using on-board cameras to determine the posture of a person by analyzing face and body bounding areas, calculating leaning values, and updating seat assignment logic to improve seat occupancy classification and safety control functions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If 2D object detection is used for seat occupancy detection, then the system can detect objects and their positions, but it cannot accurately determine person posture and leading to wrong seat assignments
Solution Approach 1:
The patent transitions from 2D object detection to 3D pose detection by introducing depth information through bounding box dimension analysis. The system calculates body bounding box dimensions (width, height, depth) and uses the aspect ratio of these dimensions to determine posture states, thereby adding a third dimension of information to resolve seat assignment ambiguities.
Solution Approach 2:
The patent changes the detection parameters from simple 2D bounding box coordinates to 3D bounding box parameters including width, height, and depth dimensions. By analyzing the aspect ratio of these dimensional parameters, the system can infer posture information and correctly assign seats even when persons are in unclear or moving positions.
2Adaptability or versatility
If a person moves or leans during detection, then the detection system may capture unclear positions, but this leads to false seat assignments
Solution Approach 1:
The patent implements a dynamic detection approach that continuously monitors bounding box dimensional changes over time. By analyzing the evolution of width, height, and depth parameters, the system can track person movement and leaning actions, adapting the seat assignment decision to the current posture state rather than relying on a single static snapshot.
Solution Approach 2:
The system uses feedback from bounding box dimension analysis to continuously refine seat assignment decisions. By monitoring changes in the aspect ratio of body bounding boxes and comparing them against expected posture patterns, the system can correct false assignments and maintain reliable seat occupancy detection even during person movement.
Data Source
AI summary
A computerized method of pose detection of a person in a vehicle includes receiving, from an on-board camera of the vehicle, an image of an interior of the vehicle showing a seat of the vehicle occupied by the person. The computerized method includes obtaining at least one first characteristic of a first face bounding area and a first body bounding area associated with the occupied seat of the vehicle. The computerized method includes determining a second body bounding area and an associated second face bounding area of the person from the image. The computerized method includes determining at least one second characteristic of the second face bounding area and the second body bounding area. The computerized method includes determining a pose of the person based on the at least one second characteristic and on the at least one first characteristic.