Dynamic Vision Sensor Corner Detection with Ordered Surface Matrices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing corner detection methods for edge IoT applications face challenges in throughput, accuracy, and energy efficiency, particularly with algorithms like LuvHarris, which are not suitable for low-power edge devices due to high energy consumption during Threshold-Ordinance Surface updates.
Innovation Solution
An adaptive corner detection method and system that organizes event data from dynamic vision sensors into a 2D array, transforms it into Ordered Surface matrices, and applies a corner detector to these matrices, optimizing memory usage and maintaining temporal ordering for efficient corner detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If LuvHarris algorithm is used to enhance processing speed and accuracy, then throughput and corner detection accuracy are improved, but energy consumption increases significantly during TOS update
Solution Approach 1:
The patent extracts the essential functionality of corner detection by processing only the Minimum Event Surface (MES) data structure rather than the complete TOS data structure. This extraction approach maintains the core corner detection capability while eliminating the energy-intensive TOS update operations, achieving a balance between processing speed and energy consumption for edge devices
Solution Approach 2:
The patent applies partial action by processing only the necessary portion of event data (MES) rather than the complete data structure (TOS). This selective processing maintains sufficient corner detection accuracy while significantly reducing computational load and energy consumption, making the algorithm suitable for resource-constrained edge devices
2Productivity
If per-event updates are applied in Asynchronous Corner Detection, then throughput is improved, but energy efficiency deteriorates for extreme-edge applications
Solution Approach 1:
The patent implements periodic action by processing event data in batches rather than continuously updating per event. Events are accumulated in a buffer and processed periodically at optimized intervals, reducing the frequency of computational operations while maintaining real-time responsiveness. This approach significantly lowers energy consumption compared to continuous per-event updates
3Loss of time
If event data is processed in real-time for corner detection, then response time is reduced, but memory requirements increase to maintain temporal ordering
Solution Approach 1:
The patent segments event data into a structured buffer format with fixed-size storage for spatial coordinates and temporal information. This segmentation allows efficient real-time processing while maintaining compact memory footprint through organized data layout and selective storage of only essential event attributes needed for corner detection
Data Source
AI summary
A method for adaptive corner detection is provided. The method includes: obtaining one or more event data from a dynamic vision sensor; capturing and organizing a plurality of recorded events of the event data into a 2D array; transforming the 2D array into one or a plurality of Ordered Surface (OS) matrices by populating one or a plurality of empty Image Matrices with the recorded events based on their coordinates and assigning order values; and applying a corner detector to the OS matrices. A system for adaptive corner detection is also provided.