Vehicle Door Control Using Gesture-Based Boarding Intention Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems struggle to accurately determine a person's intention to board a vehicle, such as a bus, due to difficulties in distinguishing between individuals waiting for the next bus or seeing someone off, particularly when focusing solely on foot position changes.
Innovation Solution
A vehicle control system that utilizes a vehicle-exterior camera and a machine learning model to analyze time-series image data, segmenting images into blocks, calculating motion vectors, and correcting their entrance-directional components to accurately detect getting-on intentions through gesture recognition, thereby controlling door operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the system focuses only on foot position changes to detect getting-on intention, then the detection process is simple, but the determination accuracy is insufficient because it cannot distinguish between persons waiting for the next bus or seeing someone off
Solution Approach 1:
The system segments the detection process into multiple independent analysis components: foot position change detection, gesture recognition, and machine learning-based intention determination. Each component processes specific features separately, improving accuracy while maintaining manageable system complexity through modular architecture
Solution Approach 2:
The system transitions from analyzing only spatial position changes (2D foot position) to incorporating temporal dimension (time-series data) and gesture dimension (hand movements, body posture). This multi-dimensional approach enables more accurate intention determination by analyzing patterns across multiple dimensions simultaneously
2Measurement precision
If the system uses only foot position data to determine getting-on intention, then the processing is fast and simple, but the system cannot accurately distinguish between different types of persons at the entrance
Solution Approach 1:
The system extracts and isolates gesture information from the overall scene data, specifically identifying hand movements, arm positions, and body orientation as separate detectable features. This extraction process captures previously lost gesture information while maintaining efficient processing through targeted feature selection
Solution Approach 2:
The machine learning model receives feedback from multiple data sources (foot position, gesture movements, temporal patterns) and continuously refines its determination of getting-on intention. The system uses the combined information from all sensors to feedback-correct the intention classification, improving accuracy by integrating multiple information streams
Data Source
Figure 1
Figure 2
Figure 3A~3B
AI summary
A vehicle control system (1) includes: a vehicle-exterior camera (3) configured to photograph a getting-on area at the vicinity of an entrance (2a); a getting-on intention determination unit (20) determining whether there is a person that is performing a predetermined getting-on intention gesture to a vehicle (2), based on time-series data of a photographed image of the vehicle-exterior camera (3), at a time of execution of a closing control of a door (25); and a door control unit (21) stopping the closing control of the door (25) and open the door (25), when the getting-on intention determination unit (20) determines that there is the person that is performing the getting-on intention gesture.