Vehicle Camera Optical Flow with Adaptive Frame Time Difference

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional methods for determining optical flow in vehicle cameras use a constant time difference, which can lead to loss of flow data in scenes with high dynamics, limiting the accuracy of self-motion estimation and scene understanding.

Innovation Solution

The method dynamically adjusts the time difference between camera images based on specific input signals, such as automatic emergency braking or pitch rate, to prevent loss of optical flow and optimize data accuracy in varying driving situations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If a constant time difference is used for optical flow calculation, then the computational process is simple, but optical flow is lost in highly dynamic scenes

Engineering Contradiction:
Improveoptical flow availabilityVSAvoidtime difference selection mechanism
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent applies dynamics by transitioning from a static constant time difference to a dynamic time difference selection mechanism. The system now adapts the time difference between consecutive images based on detected driving situations (e.g., emergency braking, high pitch rate) to maintain optical flow availability in highly dynamic scenes while keeping it constant in stable conditions.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent changes the parameter of time difference from a fixed constant to a variable parameter that is adjusted based on input signals from sensors (acceleration, pitch rate, roll rate). This allows the system to optimize optical flow calculation by selecting appropriate time differences for different driving conditions.

Inventive Principle:
Principle #35Parameter changes

2Measurement precision

If a long time difference is used, then self-motion estimation accuracy is improved, but optical flow is lost in highly dynamic situations

Engineering Contradiction:
Improveself-motion estimation accuracyVSAvoidoptical flow availability
Core Design Contradiction:
Measurement precisionVSReliability

Solution Approach 1:

The system dynamically adjusts the time difference parameter based on the current driving situation. In stable driving conditions, a longer time difference is used to improve self-motion estimation accuracy. In highly dynamic situations (detected via acceleration, pitch rate, or roll rate thresholds), the system switches to shorter time differences to prevent optical flow loss, thus maintaining both accuracy and reliability adaptively.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The time difference parameter is changed from a fixed long value to a variable that is selected from multiple possible values based on real-time sensor data. This allows the system to optimize the balance between measurement precision and reliability by adjusting the parameter according to scene dynamics.

Inventive Principle:
Principle #35Parameter changes

3Reliability

If a short time difference is used, then optical flow is maintained in dynamic scenes, but self-motion estimation accuracy is reduced

Engineering Contradiction:
Improveoptical flow availabilityVSAvoidself-motion estimation accuracy
Core Design Contradiction:
ReliabilityVSMeasurement precision

Solution Approach 1:

The system uses a dynamic time difference selection strategy that switches between short and long time differences based on driving conditions. Short time differences are used during high-dynamic events (emergency braking, sharp pitching/rolling) to maintain optical flow availability, while long time differences are used during stable driving to maximize self-motion estimation accuracy.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The time difference parameter is adjusted based on input signals from vehicle sensors. When dynamic thresholds are exceeded (indicating high-dynamic scenes), the parameter is changed to a shorter value to preserve optical flow. When thresholds are not exceeded, the parameter is set to a longer value to improve estimation accuracy.

Inventive Principle:
Principle #35Parameter changes

4Productivity

If the frame rate is dynamically changed, then computational effort is optimized, but the system complexity increases

Engineering Contradiction:
Improvecomputational efficiencyVSAvoidframe rate control mechanism
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The system dynamically changes the frame rate based on the detected driving situation and the selected time difference. This allows the system to optimize computational effort by adjusting the frame rate to match the actual dynamics of the scene, processing images at higher rates during dynamic events and lower rates during stable conditions.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The frame rate parameter is changed dynamically based on the time difference selection and sensor input signals. This adaptive parameter change allows the system to balance computational efficiency with the need for accurate optical flow calculation in varying driving conditions.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3655299B1Method and device for determining an optical flow on the basis of an image sequence captured by a camera of a vehicle
Publication Date: 2024.03.27 ROBERT BOSCH GMBH
  • EP3655299B1 patent drawingFigure 1~2
  • EP3655299B1 patent drawingFigure 3~4
  • EP3655299B1 patent drawingFigure 5

AI summary

The invention relates to a method for determining an optical flow on the basis of an image sequence captured by a camera (104) of a vehicle (100). At least one input signal (108), which represents a vehicle environment detected by using at least one sensor (110) of the vehicle (100) and/or a driving situation of the vehicle (100) and/or a result of a prior determination of the optical flow, and an image signal (106) representing the image sequence are received. A time difference value is determined by using the input signal (108). At least two individual images of the image sequence that are offset to each other by the time difference value are selected by using the image signal (106). Finally, corresponding pixels are detected in the individual images in order to determine the optical flow by using the corresponding pixels.