Mono Vision Road Profiling for Vehicle Distance and Width Estimation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for estimating the distance and width of a vehicle in front of an ego car using a monocular camera are prone to errors when assumptions about vehicle width and road flatness are violated.

Innovation Solution

A method that calculates the road profile from a video sequence using the camera's mounting position and angles, wheel suspension offsets, and visual odometry, without relying on external sensors, to accurately estimate the distance and width of a vehicle by projecting image pixels onto the calculated road profile.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If machine learning is used to train a network to estimate distance and width, then estimation accuracy improves, but device complexity and computational requirements increase

Engineering Contradiction:
Improvedistance and width estimation accuracyVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent replaces complex machine learning systems with a geometric computation approach based on camera motion profiles and road profile calculations. Instead of using trained neural networks to estimate distance and width, the system uses mathematical projections of pixel coordinates onto the calculated road profile, substituting mechanical/computational complexity with geometric principles.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Ease of manufacture

If assumptions about typical vehicle width and flat road are made, then computation is simplified, but measurement accuracy deteriorates when assumptions are violated

Engineering Contradiction:
Improvecomputation simplicityVSAvoiddistance and width estimation accuracy
Core Design Contradiction:
Ease of manufactureVSMeasurement precision

Solution Approach 1:

The patent makes the road profile dynamic by calculating it from actual camera motion data rather than assuming a static flat road. The road profile is continuously updated based on the camera's movement through the video sequence, allowing the system to adapt to varying road conditions while maintaining computational efficiency through geometric projections.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP3789908B1Mono vision system and method for a motor vehicle
Publication Date: 2025.08.27 QUALCOMM AUTO LTD
  • EP3789908B1 patent drawingFigure 1
  • EP3789908B1 patent drawingFigure 2~3

AI summary

A mono vision system (10) for a motor vehicle comprises a mono camera (12) adapted to capture images from a surrounding of the motor vehicle, and a processing device (14) adapted detect object candidates in the surrounding of the motor vehicle by processing images captured by said mono camera (12). The processing device (14) is adapted to perform the following processing with respect to a particular detected object candidate: from a plurality of images spanning a period of time between a time frame m where the object candidate (36) is visible by the camera (12) and a time frame n where the position where the object (31) corresponding to the object candidate (36) was at time m has been reached or passed by the motor vehicle (30), calculate a camera (12) motion profile C(x) between time frame m and time frame n using a visual odometry algorithm; and after time frame n, calculate a road profile R(x) between time frames m and n from said calculated camera (12) motion profile C(x).