Vehicle Range Estimation Using Ground Plane Feature Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current driver assistance systems in vehicles lack an effective method to accurately estimate the range of moving objects, such as pedestrians, in the vehicle's vicinity, which is crucial for preventing collisions during maneuvers like parking.
Innovation Solution
A method that captures images using vehicle cameras, processes features to determine a bearing to a moving object, and estimates its range based on unoccluded ground features along that bearing, utilizing camera calibration, translation, and rotation calculations to create a three-dimensional point cloud and segment features between static and moving objects, with the range estimated from the maximum distance of unoccluded ground features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If driver assistance systems use conventional camera-based detection methods, then the system can detect moving objects in the vehicle's surrounding, but the accuracy of range estimation is insufficient for collision prevention during maneuvers
Solution Approach 1:
The patent transitions from 2D image plane detection to 3D spatial reasoning by introducing ground plane modeling and depth estimation. The system projects detected moving objects onto a 3D ground plane model, using the relationship between image coordinates and real-world 3D coordinates to accurately estimate range. This dimensional transformation enables precise distance measurement by leveraging the geometric constraints of the ground plane, directly resolving the range estimation accuracy issue.
Solution Approach 2:
The patent introduces an essential matrix as an intermediary computational tool that captures the geometric relationship between two camera views. This matrix serves as a mediator that translates image coordinate differences into meaningful 3D spatial information, enabling accurate range estimation without requiring direct 3D sensing hardware. The essential matrix bridges the gap between 2D image data and 3D range information.
2Productivity
If the vehicle moves at higher speeds, then productivity of the driving maneuver increases, but the accuracy of range estimation deteriorates due to motion blur and rapid changes in scene geometry
Solution Approach 1:
The patent performs preliminary calibration to establish the essential matrix and ground plane model before actual range estimation occurs. By pre-computing the geometric relationships and transformation matrices, the system can quickly estimate ranges during high-speed maneuvers without real-time computation delays. This preliminary setup enables the system to maintain accuracy even when the vehicle moves rapidly through the scene.
Solution Approach 2:
The patent replaces traditional mechanical or hardware-based distance measurement systems with a computational approach using essential matrices and ground plane modeling. This substitution allows the system to achieve accurate range estimation through software processing rather than relying on physical sensors that may struggle with high-speed motion, thereby maintaining measurement precision during fast maneuvers.
3Measurement precision
If the system processes features from multiple captured images to determine bearing and range, then measurement precision improves, but device complexity increases due to additional processing steps
Solution Approach 1:
The patent segments the complex processing task into distinct modular stages: feature detection in individual images, feature matching across images, essential matrix computation, ground plane modeling, and final range estimation. Each module handles a specific aspect of the problem independently, making the overall system more manageable and computationally efficient despite the multi-step process required for high precision.
Data Source
Figure 1~2
Figure 3
AI summary
A method for estimating a range of a moving object (MO) comprising the steps of capturing (S1) images of a surrounding by a camera (2), processing (S2) features of captured images to determine a bearing of a moving object (MO) on the basis of a detected cluster of features belonging to the moving object (MO) within the captured images and estimating (S3) a range of the moving object (MO) on the basis of determined ground features belonging to a ground plane (GP) along the determined bearing of the moving object (MO) which are not occluded by the moving object (MO).