Dynamic Vision Sensor-Guided Image Compression for Neural Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Neural networks in safety-critical systems like autonomous vehicles require accurate and rapid classification probabilities, which are challenging due to the need to process moving obstacles in real-time.
Innovation Solution
A method involving a Dynamic Vision Sensor and an image sensor capturing an overlapping field of view, applying higher image compression to areas of no movement to enhance accuracy and speed of classification by a neural network.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If higher image compression is applied to reduce processing time, then processing speed improves, but classification accuracy deteriorates
Solution Approach 1:
The patent applies different compression levels to different regions of the image based on movement detection. Areas with movement (potentially containing obstacles) are compressed at lower levels to preserve detail, while static areas are compressed at higher levels to reduce overall data volume. This local differentiation resolves the contradiction by maintaining accuracy where needed while improving overall processing speed.
Solution Approach 2:
The image is segmented into multiple regions based on movement detection results. The processing circuitry divides the image into areas with movement and areas without movement, applying different compression strategies to each segment. This segmentation allows the system to optimize processing speed for static regions while preserving classification accuracy for dynamic regions containing potential obstacles.
2Productivity
If image compression is increased to process more images per second, then real-time processing capability improves, but signal-to-noise ratio deteriorates
Solution Approach 1:
Different compression levels are applied to different image regions based on their importance. Regions containing movement (potential obstacles) maintain higher signal-to-noise ratios through lower compression, while static background regions accept higher compression with minimal impact on classification accuracy. This enables processing more images per second while preserving critical information.
3Loss of time
If uniform high compression is applied to the entire image, then processing time is reduced, but accuracy in movement detection areas deteriorates
Solution Approach 1:
The system applies local quality differentiation by detecting movement regions and applying appropriate compression levels to each. Movement detection areas receive lower compression to maintain reliability for safety-critical decisions, while static areas receive higher compression to reduce overall processing time. This resolves the contradiction between processing speed and decision accuracy.
Solution Approach 2:
The system performs preliminary movement detection and region classification before applying compression. By identifying which areas contain movement in advance, the system can pre-determine the appropriate compression level for each region, ensuring that critical areas maintain accuracy while non-critical areas are efficiently compressed, thereby reducing overall processing time without sacrificing reliability.
Data Source
AI summary
A method comprising the steps of: receiving an image from a dynamic vision sensor and an image from an image sensor, wherein the dynamic vision sensor and the image sensor capture at least an overlapping field of view; determining, from the image received from the dynamic vision sensor, an area of movement in the field of view of the image from the image sensor; applying a higher level of image compression to the areas of no movement compared with the area of movement in the image from the image sensor to produce a processed image; and outputting the processed image to a neural network.