Real-time Moving Object Detection via Grid-based Motion Vector Residuals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current advanced drive assist systems (ADAS) and vehicle-mounted cameras lack real-time detection of moving objects around the camera, which is essential for safe driving and surveillance applications.
Innovation Solution
An image processing method and device that constructs grids on an image, extracts features, estimates motion vectors, and detects moving objects by calculating residual vectors between first and second motion vectors, with filtering techniques to remove noise and identify candidate moving objects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If real-time moving object detection is implemented, then detection capability is improved, but calculation time and processing complexity increase
Solution Approach 1:
The image is divided into multiple grids, and motion analysis is performed independently for each grid. This segmentation allows parallel processing of different regions, reducing overall calculation time while maintaining detection capability across the entire image.
Solution Approach 2:
The system calculates motion vectors only for feature points within each grid rather than processing all pixels. This partial action approach significantly reduces computational load while still providing accurate moving object detection through residual vector analysis.
2Measurement precision
If motion vectors are estimated for all features, then detection accuracy is improved, but computational complexity increases
Solution Approach 1:
Feature points are grouped into discrete grids, and motion vector estimation is performed separately for each grid. This segmentation reduces the overall computational complexity by breaking down a large-scale problem into smaller, manageable sub-problems that can be processed independently.
Solution Approach 2:
Motion vectors are estimated only for feature points that fall within each grid rather than attempting to process all possible points. This partial action maintains detection accuracy for moving objects while significantly reducing processing complexity through selective feature analysis.
3Measurement precision
If residual vectors are calculated for all features, then moving object detection accuracy is improved, but processing time increases
Solution Approach 1:
The calculation of residual vectors is performed separately for each grid containing feature points. This segmentation enables parallel processing of residual vector calculations across different image regions, improving processing speed while maintaining detection accuracy through comprehensive residual analysis.
Solution Approach 2:
Residual vectors are calculated only for feature points within identified grids rather than for all pixels in the image. This partial action approach maintains high detection accuracy by focusing computational resources on relevant feature points while significantly improving overall processing speed.
Data Source
AI summary
Provided are an image processing system and an image processing method. An image processing method includes: constructing a plurality of grids on an image; extracting features of the image; estimating a grid motion vector of each of the plurality of grids, based on a first motion vector of each of the features; estimating a second motion vector of each feature based on the grid motion vector; and detecting a moving object based on a difference between the first motion vector and the second motion vector.