Autonomous Parking Control Using Wheel-Based Ground Clearance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing autonomous parking systems using ultrasonic signals and cameras face errors in determining the ground clearance of surrounding vehicles, leading to inaccuracies in distance measurement, with spatial recognition methods like Semantic Segmentation often overestimating or underestimating the distance due to incomplete vehicle representation.
Innovation Solution
A vehicle system that combines spatial recognition and object recognition using camera data to accurately determine ground clearance by analyzing wheel area data, bumper area data, and pedestrian data, correcting the distance measurement by reflecting the ground clearance in the remaining area excluding the wheel area, and converting key points into world coordinates for precise positioning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If spatial recognition using Semantic Segmentation is used to label the vehicle in its visible form, then the vehicle can be recognized in the image, but the ground clearance cannot be accurately determined resulting in error in distance measurement
Solution Approach 1:
The patent applies segmentation by dividing the vehicle recognition into two distinct parts: spatial recognition (vehicle body) and object recognition (wheel area). This allows the system to separately process and measure the ground clearance based on wheel area data, thereby resolving the contradiction between vehicle recognition and accurate ground clearance measurement.
2Reliability
If the vehicle is labeled in its visible form using Semantic Segmentation, then the visible areas are recognized, but the distance to the actual vehicle is overestimated
Solution Approach 1:
The patent introduces wheel area data as an intermediary element that mediates between the visible vehicle form and the actual distance measurement. By using the wheel area (which contacts the ground) as a reference point, the system can accurately determine the ground clearance and correct the distance measurement, thereby improving both reliability and precision.
3Device complexity
If ground clearance is determined using spatial recognition data alone, then the process is simple, but the distance to surrounding vehicles is inaccurate
Solution Approach 1:
The patent merges spatial recognition data and object recognition data (wheel area data) to determine ground clearance. This combination allows the system to maintain relatively simple processing while achieving accurate distance measurement, as the wheel area data provides the necessary ground contact reference without requiring complex additional sensors or processing.
Data Source
AI summary
A vehicle may include a camera obtaining a surrounding image around the vehicle; and a controller configured to derive spatial recognition data by learning the surrounding image of the vehicle as an input value of the controller, derive object recognition data including wheel area data of surrounding vehicles around the vehicle by learning the surrounding image of the vehicle as an input value of the controller, determine a ground clearance between a bottom surface of a vehicle body of the surrounding vehicles and a ground by use of the spatial recognition data and the wheel area data, and control the vehicle to park the vehicle according to the ground clearance.