Trailer Edge Tracking via Geometric Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems for tracking an image section on a monitor to replace an exterior mirror in vehicles, particularly when cornering with a trailer, face challenges in reliably detecting the trailer's rear edge due to motion detection difficulties and computational limitations, leading to undesirable adjustments in the framing.
Innovation Solution
A method that determines a three-dimensional model of the scene, calculates movement vectors, and estimates the most probable position of the trailer's rear edge using Bayesian estimation, incorporating variance values and assumptions about the trailer's geometry, to robustly track the image section with low computational effort.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If motion detection is used to track the trailer edge, then the image section can be adjusted to follow the trailer, but reliable motion detection is difficult to implement and leads to undesirable adjustments
Solution Approach 1:
The patent replaces complex motion detection algorithms with a geometric model-based approach. Instead of detecting motion vectors from image sequences, the system calculates the expected trailer edge position based on vehicle and trailer geometry models, transforming the problem from image processing to geometric computation.
Solution Approach 2:
The patent changes the approach from detecting motion parameters (velocity, acceleration) to using geometric parameters (vehicle position, trailer angle, coupling point location). This parameter transformation simplifies the tracking problem by using readily available vehicle state data instead of complex image-based motion analysis.
2Ease of operation
If the image section is tracked to keep the trailer edge in view, then the driver can see the trailer during cornering, but driver head movements cause undesirable framing adjustments
Solution Approach 1:
The system uses feedback from the geometric model to distinguish between intentional trailer position changes and driver head movements. By comparing the modeled trailer edge position with actual image features, the system can determine whether framing adjustments are necessary or if they result from driver head movement, thereby maintaining framing stability.
Solution Approach 2:
The geometric model acts as an intermediary between the camera image and the display framing. Instead of directly reacting to image changes, the system uses the model to filter and interpret position changes, determining whether they represent actual trailer movement requiring framing adjustment or driver head movement that should be ignored.
3Reliability
If a three-dimensional model with Bayesian estimation is used to determine the trailer edge position, then tracking reliability is improved, but computational effort increases
Solution Approach 1:
The patent applies Bayesian estimation selectively rather than comprehensively. Instead of performing full Bayesian analysis on all image features, the system uses it only for determining the trailer edge position from the geometric model, combining it with simpler image processing for other aspects, thus reducing overall computational energy while maintaining accuracy where it matters most.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to a method for tracking a section of a camera image (104) displayed on a monitor (101) to replace an exterior mirror (102) of a vehicle (103) during cornering of a vehicle combination (106) consisting of the vehicle (103) and a trailer (105), as well as a corresponding control unit (110). First, a three-dimensional model (107) is determined from the camera image (104, KB) with the object position (P3, P5, P8) of at least one object (103, 105, 108) and predefined modeling parameters (MP1, MP2, ...). Then, a motion vector (1BV5, 2BV5) to at least one distinctive point (KP1) of the trailer (105) is determined from the camera image (104). The most probable position (PK) of point (KP1) is determined from the object position (P5) and the determination of the motion vector (1BV5, 2BV5). Then the image section (104a, 104b) is adjusted according to the most probable position (PK) of point (KP1).