Video Group Behavior Analytics Using Keypoint Time Series
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video analytics systems struggle to effectively identify and analyze the behaviors of different groups within video data, particularly in multi-camera surveillance systems, making it challenging to detect anomalous situations such as a child becoming lost or kidnapped.
Innovation Solution
A device represents objects in video data as a set of timeseries of key points, forms groups based on these timeseries, and uses self-supervised learning to identify anomalous behaviors, providing alerts through a user interface.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional video analytics methods are used to identify groups and analyze behaviors, then the system complexity remains manageable, but the ability to detect anomalous group behaviors is insufficient
Solution Approach 1:
The patent segments the video analysis process into distinct components: object detection, key point extraction, timeseries generation, group formation, and anomaly detection. This segmentation allows each component to be optimized independently while maintaining overall system manageability, resolving the contradiction between detection accuracy and system complexity.
Solution Approach 2:
The patent transforms spatial video data into temporal timeseries representations by extracting key points across multiple frames. This dimensionality change from spatial to temporal domain enables more effective anomaly detection through timeseries analysis, improving detection accuracy while providing a structured approach to manage complexity.
2Loss of information
If detailed analysis of individual objects is performed, then the understanding of object behaviors improves, but the computational resources and processing time increase
Solution Approach 1:
The patent extracts only the essential key points from full object data, selecting critical features that capture behavioral information while discarding redundant details. This extraction approach maintains information about object behaviors while significantly reducing processing time and computational resources.
Solution Approach 2:
The patent transforms detailed spatial coordinates into simplified timeseries parameters that capture the essence of object behavior. By changing the representation parameters from raw coordinates to aggregated timeseries features, the system retains behavioral information while reducing processing complexity and time.
3Area of stationary object
If multiple video streams are analyzed simultaneously, then the coverage area increases, but the difficulty of identifying and comparing group behaviors across streams increases
Solution Approach 1:
The patent creates a universal timeseries representation that can be applied across multiple video streams independently. This universal format enables consistent group identification and behavior comparison across different streams, making the analysis scalable to larger coverage areas without proportionally increasing detection difficulty.
Data Source
AI summary
In one embodiment, a device represents each of a plurality of objects depicted in video data over time as a set of timeseries of key points associated with that object. The device forms groups of objects based on their associated sets of timeseries. The device identifies an anomalous behavior of a particular group from among the groups of objects, based on their constituent timeseries of coordinate points. The device provides an alert regarding the anomalous behavior to a user interface for review.