Dynamic Vision Sensor Architecture for Event-Based Change Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional frame-based image sensors in machine vision systems suffer from high data redundancy, limited dynamic range, poor low-light performance, motion blur, and computational complexity in solving the correspondence problem, leading to increased power consumption and reduced reaction time.
Innovation Solution
A Dynamic Vision Sensor (DVS) design with synchronous pixel operation, controlled time resolution, reduced pixel size, and adaptive event rate, utilizing a single comparator per pixel for efficient change detection, and a global shutter mechanism to minimize motion artifacts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If conventional frame-based image sensors are used, then complete scene information is captured, but data redundancy increases significantly
Solution Approach 1:
The patent extracts only the essential information from the scene by using change detection pixels that respond only to changes in light intensity. Instead of capturing complete frame data, the sensor extracts and transmits only the changes, removing redundant information while preserving important scene dynamics.
Solution Approach 2:
The patent implements different pixel types with specialized functions within the same sensor array. Change detection pixels are distributed throughout the sensor, each locally detecting changes in its specific region. This local specialization allows efficient change detection without requiring complete frame capture.
2Speed
If high temporal resolution is achieved through frequent frame capture, then motion tracking improves, but power consumption increases
Solution Approach 1:
The patent uses periodic reset signals applied to the change detection pixels to enable continuous monitoring without continuous data transmission. The pixels integrate changes between reset periods, allowing high temporal resolution measurement capability while consuming power only during reset and readout phases rather than continuously.
3Measurement precision
If pixel circuits include amplifier and two comparators for change detection, then detection precision improves, but pixel area increases
Solution Approach 1:
The patent combines the amplifier and comparator functions into a single integrated circuit block within each pixel. By merging these functions and using shared circuitry, the design achieves precise change detection capability while reducing the overall pixel area compared to having separate amplifier and dual comparator circuits.
4Loss of time
If asynchronous readout is used for high temporal resolution, then reaction time improves, but timing jitter increases
Solution Approach 1:
The patent incorporates timestamp generation with feedback mechanisms that record precise timing information for each detected change event. The system uses synchronized clock signals and timestamp counters that provide feedback on timing, allowing asynchronous event detection while maintaining accurate timing information for post-processing and analysis.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The DVS achieves high temporal resolution, low power consumption, and efficient data processing with reduced motion blur, overcoming the limitations of conventional cameras by providing smart data for computer applications.
Implementation Method 1
A dynamic vision sensor (DVS) or change detection sensor reacts to changes in light intensity
Data Source
AI summary
A dynamic vision sensor (DVS) or change detection sensor reacts to changes in light intensity and in this way monitors how a scene changes. This disclosure covers both single pixel and array architectures. The DVS may contain one pixel or 2-dimensional or 1-dimensional array of pixels. The change of intensities registered by pixels are compared, and pixel addresses where the change is positive or negative are recorded and processed. Analyzing frames based on just three values for pixels, increase, decrease or unchanged, the proposed DVS can process visual information much faster than traditional computer vision systems, which correlate multi-bit color or gray level pixel values between successive frames.