Overlapping Time Windows for Event-Based Vision Sensor Frame Accumulation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional computer vision methods struggle to effectively process the event stream from event-based vision sensors due to the absence of frames, particularly for applications involving continuously moving objects.

Innovation Solution

The method involves using overlapping time windows to accumulate events into image frames, where the first and second image frames are elaborated using events from overlapping time windows, improving the precision of subsequent computer vision algorithms.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional computer vision methods are used to process event stream data, then processing speed is maintained, but precision is degraded due to the absence of frames and large differences in events between non-overlapping time windows

Engineering Contradiction:
Improveprecision of computer vision algorithmsVSAvoiddifferences in events between time windows
Core Design Contradiction:
Measurement precisionVSLoss of information

Solution Approach 1:

The event stream is segmented into multiple overlapping time windows instead of non-overlapping windows. Each time window captures events with temporal overlap, ensuring that transitions and movements are captured in multiple consecutive frames. This segmentation strategy reduces information loss at window boundaries and improves precision for detecting fast-moving objects.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The method performs preliminary accumulation of events in overlapping time windows before applying computer vision algorithms. By pre-processing the event stream with overlapping windows, the system prepares enhanced frame data that preserves temporal continuity, thereby improving subsequent algorithm precision without requiring changes to the algorithms themselves.

Inventive Principle:
Principle #10Preliminary action

2Productivity

If non-overlapping time windows are used to accumulate events, then device complexity is reduced, but productivity is degraded due to rigid and discontinuous frame integration

Engineering Contradiction:
Improvecontinuous motion detection capabilityVSAvoidcomplexity of time window management
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The system dynamically manages overlapping time windows by continuously accumulating events in sliding temporal intervals. This dynamic approach allows frames to be generated continuously with temporal overlap, improving productivity for continuous motion detection. The complexity is managed through systematic event buffering and timestamp-based filtering rather than complex hardware modifications.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The overlapping time windows enable continuous accumulation of events without gaps between frames. This continuity ensures that no motion events are missed at window boundaries, maintaining productive continuous monitoring. The method keeps the data accumulation process ongoing by overlapping windows, thereby eliminating idle periods between frame generations.

Inventive Principle:
Principle #20Continuity of useful action

3Adaptability or versatility

If fixed integration time is used for event accumulation, then device complexity is minimized, but adaptability is reduced for varying motion speeds and conditions

Engineering Contradiction:
Improveadaptability to varying motion speedsVSAvoidcomplexity of adaptive time window management
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The system changes the temporal parameter of time windows by using overlapping intervals that can be adjusted based on motion characteristics. This parameter change allows the system to adapt to varying motion speeds - faster motion can be captured with shorter or more overlapping windows, while slower motion uses longer intervals. The adaptability is achieved through flexible time window configuration rather than hardware changes.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP4060983B1A method for accumulating events using an event-based vision sensor and overlapping time windows
Publication Date: 2025.04.09 TOYOTA JIDOSHA KK
  • EP4060983B1 patent drawingFigure 1
  • EP4060983B1 patent drawingFigure 2
  • EP4060983B1 patent drawingFigure 3

AI summary

A method for elaborating image frames using an event-based vision sensor (PM) which comprises a plurality of pixels and is configured to output a stream of change events (ES), wherein each change event corresponds to a change in brightness detected by one pixel and comprises a time stamp, a pixel address, and a parameter describing a sign of said brightness change, wherein a first image frame and a second image frame are elaborated, using change events accumulated in buffers, these change events appearing in time windows which have a relative time position. The invention also concerns the corresponding system.