Video Object Indexing System for Multi-Camera Stream Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video analytic systems require significant operator time to monitor and analyze video frames from multiple cameras, as they lack efficient methods to identify and track objects of interest across different video streams, leading to inefficiencies in locating and displaying relevant video frames.
Innovation Solution
A method and system that perform a 'seed search' to identify video objects matching a selected object, followed by a 'complete search' to find matching objects across all video data, allowing for the display of a subset of video frames containing these matches, thereby reducing the need for operators to view non-relevant frames.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If current video analytic systems display all video frames from multiple cameras for monitoring, then operators can locate objects of interest, but the time required to monitor and analyze video frames increases significantly
Solution Approach 1:
The system performs preliminary actions by pre-processing video frames to detect objects of interest and generating analytical data before operator review. Motion detection and object identification are carried out in advance, so that when operators need to locate specific objects, the system has already prepared indexed information about detected objects across multiple cameras, eliminating the need for operators to manually scan all video frames.
Solution Approach 2:
The patent introduces an intermediary indexing system that mediates between raw video data and operator analysis. The analytical data structure serves as an intermediary layer that organizes video frame information by detected objects, cameras, and timestamps. This intermediary structure allows operators to query and retrieve relevant video frames efficiently without manually monitoring all streams, thus reducing monitoring time while maintaining detection accuracy.
2Reliability
If operators monitor all video frames from multiple cameras to find objects of interest, then complete coverage is achieved, but the complexity of the monitoring system increases
Solution Approach 1:
The patent applies segmentation by dividing the monitoring task into distinct functional components: motion detection modules for each camera, object identification algorithms, analytical data generation, and video frame retrieval systems. The analytical data structure itself is segmented into multiple fields (object identifiers, camera identifiers, timestamps, confidence scores) that can be independently processed and queried. This segmentation allows the system to maintain comprehensive monitoring coverage while managing complexity through modular architecture.
3Productivity
If motion detection analytics are used to display only frames with moving objects, then the number of frames to monitor is reduced, but the system cannot find objects that appear stationary or are captured by different cameras
Solution Approach 1:
The patent implements universality by designing an analytical data structure and search system that can handle multiple types of queries beyond motion detection. The system can search for objects by identifier across any camera, retrieve frames within specific time ranges, and locate objects regardless of whether they are in motion. The analytical data structure is designed to be universally applicable to various search criteria, making the system versatile for different monitoring needs while maintaining efficient processing through pre-generated indexes.
Data Source
AI summary
A seed search of a subset of analytical data corresponding to video objects displayable in a plurality of video frames is carried out to identify video objects that most closely match a selected video object and then complete searches of the analytical data may be carried out so as to identify video objects that most closely match each video object identified during the seed search. The video objects having the greatest number of occurrences of being identified during the complete searches may be displayed by a graphical user interface (GUI). In this way, the GUI may display the video objects in an order based on how closely each video object matches the selected video object and/or a video object identified during the seed search, which may an order different than an order based on a time when each video object was captured.