Digital Neuromorphic Vision System for High Frame Rate Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image processing systems face challenges in efficiently analyzing high frame rate video data due to overwhelming computational demands, as they process all frames without prioritizing changes, leading to limited ability in real-time object detection and tracking.
Innovation Solution
A digital Neuromorphic (NM) vision system that uses a digital retina and engine to generate spike data based on changes in image frames, incorporating CMOS technology and velocity transformation modules to extract and encode spatio-temporal differences, enabling improved object detection, classification, and tracking.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If conventional image processing systems process all frames at high frame rate, then temporal resolution is improved, but computational demand and data processing load increase significantly
Solution Approach 1:
The patent extracts only the essential information from video frames by detecting changes between frames. Instead of processing all pixel data in every frame, the system identifies and processes only regions where changes occur, thereby maintaining high temporal resolution while significantly reducing computational demand.
Solution Approach 2:
The system performs partial processing by focusing computational resources only on changed regions rather than processing the entire frame. This partial action approach maintains the ability to detect fast movements while reducing overall computational load by avoiding processing of unchanged areas.
2Reliability
If all frame data is processed for object detection, then detection completeness is improved, but processing time and throughput are reduced
Solution Approach 1:
The patent extracts only the necessary information for object detection by identifying changed regions between frames. This extraction approach maintains detection completeness for moving objects while improving processing throughput by eliminating redundant processing of static background areas.
Solution Approach 2:
The system segments the processing task by dividing the frame into changed and unchanged regions. Only the changed regions undergo detailed object detection processing, while unchanged regions are quickly identified and skipped, thereby maintaining detection completeness while significantly improving processing throughput.
3Measurement precision
If high spatial and temporal resolutions are maintained in video processing, then object detection accuracy is improved, but data volume and processing requirements increase
Solution Approach 1:
The patent extracts only the essential data needed for accurate object detection by identifying changed regions. This approach maintains high measurement precision for moving objects while reducing data volume by not storing or processing unchanged background data.
Solution Approach 2:
The system applies local quality by maintaining high processing resolution only in changed regions where objects are likely to be present, while using coarser or no processing in unchanged regions. This preserves object detection accuracy in critical areas while reducing overall data volume.
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 digital NM vision system reduces data processing load by focusing on frame differences, enhancing accuracy and throughput in object detection and tracking, while maintaining high spatial and temporal resolutions without the need for costly analog circuitry in each pixel.
Implementation Method 1
a detector that includes a photodetector array that converts rays of light into image data
Data Source
AI summary
A system and methodologies for neuromorphic vision simulate conventional analog NM system functionality and generate digital NM image data that facilitate improved object detection, classification, and tracking.