Grid-Based Motion Pattern Model for Video Deviation Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional motion detection algorithms in video surveillance are computationally heavy, leading to high hardware costs and inability to distinguish between normal and abnormal motion, necessitating a more efficient method for detecting deviation from a motion pattern.
Innovation Solution
A grid-based motion pattern model is used to detect deviation, which reduces storage and computational requirements by processing motion pattern grid data instead of raw video data, allowing for efficient detection and analysis of motion patterns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional image processing algorithms are used for motion detection, then motion detection capability is achieved, but hardware costs and computing resources become too high
Solution Approach 1:
The patent extracts only the essential motion information from video frames by comparing current frames with previous frames and generating a motion pattern grid that represents abnormal motion. This extracted motion pattern grid is then used for deviation detection, eliminating the need to process entire video frames and significantly reducing hardware and computing resource requirements.
Solution Approach 2:
The patent segments the video analysis process into two stages: first generating motion pattern grids from video frames, then using these grids for deviation detection. This segmentation allows the system to process only the essential motion data rather than entire video streams, reducing computational complexity and hardware costs.
2Reliability
If conventional motion detection algorithms are used, then motion detection is performed, but the ability to distinguish between normal and abnormal motion is lost
Solution Approach 1:
The patent performs preliminary action by generating motion pattern grids that capture the expected motion patterns during a training phase. These pre-computed motion pattern grids serve as reference models that enable the system to distinguish between normal motion (matching the pattern) and abnormal motion (deviating from the pattern), providing the precision needed to differentiate between the two.
3Loss of information
If complete video recordings are transferred for analysis, then comprehensive data is available, but network capacity requirements become excessive
Solution Approach 1:
The patent extracts only the motion pattern grid data from complete video recordings, which contains the essential information needed for deviation detection. This extracted motion pattern grid is then transferred and analyzed instead of the entire video recording, dramatically reducing network capacity requirements while preserving the necessary information for accurate deviation detection.
Solution Approach 2:
The system performs preliminary processing of video data to generate motion pattern grids during the training phase. These pre-processed motion pattern grids are then stored and used for subsequent deviation detection, eliminating the need to transfer and analyze complete video recordings and significantly reducing network bandwidth requirements.
4Device complexity
If motion pattern grid data is stored and used for deviation detection, then computing resources are optimized, but storage space requirements must be considered
Solution Approach 1:
The patent extracts motion pattern grid data from video frames, which represents only the essential motion information. This extracted motion pattern grid is significantly smaller in size compared to the original video data, reducing storage space requirements while enabling efficient deviation detection with optimized computing resources.
Data Source
AI summary
A current motion grid comprising a plurality of elements is generated by storing in each element of the current motion grid an indication of whether there is a change between corresponding elements of at least two images captured from a video sequence. A current motion pattern grid comprising a plurality of elements is generated by firstly searching for a segment consisting of grid elements in which a change has been indicated in the current motion grid and which are neighbouring to one another and, secondly, storing in each element of the segment a value corresponding to a size of the segment. A value of an element of the current motion pattern grid is compared with a threshold value. It is then determined, based on the result of the comparison, whether there is deviation from the motion pattern.