Video Sampling Using Spatiotemporal Constraints
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video surveillance systems fail to efficiently reduce the size of stored video data by selectively removing segments that do not contain track sequences of people, relying on event-based or content-based approaches that do not apply spatiotemporal constraints to track sequences for statistical behavior analysis.
Innovation Solution
A method and system that select and store video segments based on semantically-meaningful and domain-specific spatiotemporal selection criteria, identifying trip information of people across multiple camera views and compacting video streams by removing segments without trip information, using a proprietary video storage format that associates video segments with trip data for efficient storage and analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If event-based or content-based approaches are used to manage video data, then video segments can be selectively stored, but the storage size reduction is insufficient and spatiotemporal constraints are not applied to track sequences
Solution Approach 1:
The video stream is divided into segments based on track sequence boundaries. The system identifies start and end frames of each track sequence and extracts only those segments, discarding the rest. This segmentation approach enables precise control over stored video content while maintaining systematic organization of the video data.
Solution Approach 2:
The system performs preliminary analysis of video streams to identify track sequences before storage decisions are made. By pre-processing the video data to detect human tracks and determine their temporal-spatial boundaries, the system can make informed decisions about which segments to store, avoiding the need for complex post-processing or retrieval operations.
2Loss of information
If all video segments are stored for comprehensive behavior analysis, then complete data is available, but storage requirements become prohibitively large
Solution Approach 1:
The system extracts only the video segments that contain track sequences of human interest from the complete video stream. By removing segments without relevant track information and segments where tracks do not satisfy spatiotemporal constraints, the system achieves up to 95% storage reduction while preserving all necessary information for statistical behavior analysis.
Solution Approach 2:
Different quality levels or storage priorities are applied to different video segments based on their content. Video segments containing track sequences that satisfy spatiotemporal constraints are retained with full quality, while segments without such sequences are removed entirely. This local differentiation optimizes storage efficiency without compromising analysis quality.
3Productivity
If spatiotemporal constraints are applied to track sequences, then storage efficiency improves, but the complexity of selecting video segments increases
Solution Approach 1:
The system dynamically adjusts the application of spatiotemporal constraints based on the detected track sequences. Rather than applying fixed, complex filtering rules to all video segments, the system adapts the constraint application to the actual content, only enforcing spatiotemporal filtering on segments that contain relevant track information. This dynamic approach simplifies the overall selection process while maintaining high storage efficiency.
Data Source
AI summary
The present invention is a method and system for selecting and storing videos by applying semantically-meaningful selection criteria to the track sequences of the trips made by people in an area covered by overlapping multiple cameras. The present invention captures video streams of the people in the area by multiple cameras and tracks the people in each of the video streams, producing track sequences in each video stream. The present invention determines a first set of video segments that contains the trip information of the people, and compacts each of the video streams by removing a second set of video segments that do not contain the trip information of the people from each of the video streams. The present invention selects video segments from the first set of video segments based on predefined selection criteria for the statistical behavior analysis. The stored video data is an efficient compact format of video segments that contain the track sequences of the people and selected according to semantically-meaningful and domain-specific selection criteria. The final storage format of the videos is a trip-centered format, which sequences videos from across multiple cameras, and it can be used to facilitate multiple applications dealing with behavior analysis in a specific domain.