Video Motion Detection via Spatio-Temporal Volume Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current motion detection methods in video analytics are sensitive to lighting changes and computationally demanding, failing to accurately distinguish between true motion and lighting effects, and are inefficient for real-time processing.
Innovation Solution
A method involving the creation of an image volume from a video sequence, sliced along the (X,t) or (Y,t) plane, followed by logarithmic transformation to separate reflectance and illumination components, spatio-temporal filtering to enhance edges, and a threshold function to identify motion, which suppresses lighting changes and enhances edge detection for efficient motion identification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If change detection methods or background modeling methods are used, then motion detection can be performed, but the methods are very sensitive to lighting changes and cannot distinguish between true image changes and lighting effects
Solution Approach 1:
The patent transforms the motion detection problem from 2D spatial image analysis to 3D spatio-temporal volume analysis by stacking multiple frames. This dimensional extension enables the use of spatio-temporal filters that can distinguish true motion from lighting changes by analyzing patterns across both space and time dimensions simultaneously
Solution Approach 2:
The patent replaces traditional mechanical/image-processing approaches (background subtraction, Gaussian modeling) with a spatio-temporal filtering approach that operates on the volume of image frames. This substitution enables more robust motion detection by leveraging temporal correlations that are invisible to frame-by-frame analysis
2Measurement precision
If motion estimation methods such as optical flow are used, then velocities for image regions in motion can be determined, but the methods are very computationally demanding and inaccurate
Solution Approach 1:
Instead of performing full optical flow computation across the entire image, the patent applies spatio-temporal filtering selectively to detect motion regions. This partial action approach achieves sufficient motion detection accuracy without the excessive computational cost of complete optical flow calculation
Solution Approach 2:
The patent segments the motion detection task into spatial filtering and temporal filtering components applied to the spatio-temporal volume. This segmentation allows each filter to operate independently and efficiently, reducing overall computational demand compared to unified motion estimation methods
3Measurement precision
If most currently available spatio-temporal filters are used, then images can be processed simultaneously spatially and temporally, but they are too computationally demanding for real-time embedded processing
Solution Approach 1:
The patent modifies the parameters of spatio-temporal filtering by applying logarithmic transformation to the image volume before filtering. This parameter change simplifies the filtering operation and enables real-time processing while maintaining spatio-temporal analysis capabilities
Solution Approach 2:
The patent uses simple, computationally inexpensive operations (logarithmic transformation, basic spatio-temporal filtering) that can be performed rapidly on embedded systems. These lightweight operations provide sufficient motion detection performance without requiring complex, resource-intensive algorithms
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method for video motion detection, a computer program and a camera combines a series of images of a video sequence into an image volume having vertices of X, Y and t. The image volume is sliced in either the (X,t) or (Y,t) planes, filtered with a spatio-temporal filter and thresholded to reduce the amount of information. Then a search algorithm searches along search lines in a subset of the video sequence to locate motion. Motion can be detected by identifying lines that are not parallel to any of the vertices. Stationary subjects and lighting changes appear as lines parallel to one of the vertices. Thus, true motion can be distinguished from lighting changes.