Video Motion Detection Using Cell Segmentation and Noise Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video capture devices face challenges in detecting very small, aperiodic motions due to pixel-level noise floors that are similar in magnitude to the motion, and they may not be capable of capturing video at high, regular frequencies.
Innovation Solution
A motion detection system that segments video images into a grid of cells, calculates difference metrics based on pixel intensity values over time, and uses sensor noise modeling to differentiate between true object motion and sensor noise, enabling the identification of motion patterns such as respiratory activity, earthquake vibrations, or seizure activity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If video capture devices use standard pixel-level noise floors for image capture, then the device complexity and manufacturing cost are reduced, but the measurement precision for detecting very small motions deteriorates because the noise floor magnitude is similar to the motion to be detected
Solution Approach 1:
The patent divides the image into multiple cells or regions of interest, allowing motion detection to be performed independently in each segment. This segmentation enables the system to detect small motions in specific areas without being overwhelmed by noise across the entire image, thereby improving measurement precision without requiring complex hardware modifications
Solution Approach 2:
The patent performs preliminary actions by capturing multiple images at different times and computing difference metrics between them before final motion determination. This preliminary processing allows the system to accumulate information about motion patterns over time, distinguishing true motion from noise through temporal analysis rather than relying on single-frame precision
2Measurement precision
If video capture devices capture video at high, regular frequencies to detect periodic motions, then the measurement precision for periodic motion detection is improved, but the adaptability to detect aperiodic motions deteriorates because the system assumes periodicity in the analysis method
Solution Approach 1:
The patent employs dynamic analysis methods that adapt to the temporal characteristics of motion without assuming periodicity. The system computes difference metrics between images at different times and analyzes the statistical properties of these differences, allowing it to detect both periodic and aperiodic motions. The analysis dynamically adjusts to the actual motion pattern rather than forcing a periodic model
Solution Approach 2:
The patent changes the parameters of motion detection by using statistical measures such as mean and standard deviation of difference metrics across multiple time points. This parameter transformation allows the system to detect motion patterns regardless of whether they are periodic or aperiodic, improving adaptability while maintaining precision through robust statistical analysis
3Measurement precision
If the video feed is processed by segmenting into cells and computing difference metrics for all pixel pairs, then the measurement precision for motion detection is improved, but the computational resources and processing time increase significantly
Solution Approach 1:
The patent segments the image into cells and computes difference metrics only for corresponding pixels across cells rather than all pixel pairs. This segmented approach reduces the computational complexity from O(N²) to O(N) where N is the number of pixels, significantly reducing processing time while maintaining precision through localized analysis
Solution Approach 2:
The patent extracts only the necessary information by computing difference metrics for pixels at the same spatial location across different cells, rather than comparing all possible pixel pairs. This extraction of relevant differences reduces the computational burden while preserving the essential motion information needed for accurate detection
Data Source
AI summary
A system for detecting motion in one or more images of a video feed captured by an image capture device. The system collects the multiple images of the video feed and segments each image into multiple portions or cells. Multiple cell groups are determined based on the images, with each cell group including information relating to contents of set of a common portion of the multiple images. A difference metric is calculated for each of the cell groups and at least one level of change of values of the difference metric is determined. The at least one level of change is data representing the video feed that is compared with a still-scene data model associated with the image capture device to determine a respective probability that each of the cell groups includes motion. Motion is identified for those cell groups that have a probability greater than or equal to a threshold value. The system generates a transformed video feed including the video feed and a visual indication of the motion.