Predictive Graph Maps for Multi-Camera Surveillance Capture
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video surveillance systems often miss detecting fast-moving objects, objects outside the current field of view, or early portions of video events due to delayed or inefficient video capture modifications, particularly when using object detection and recognition features.
Innovation Solution
A system that uses graph maps to predictively modify video capture operations across networked cameras by sharing video data to adjust field of view and capture rates based on spatial relationships, allowing cameras to prepare for object entry before it occurs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If video cameras use trigger conditions to selectively capture high quality video, then storage cost is reduced, but object detection is delayed and fast moving objects are missed
Solution Approach 1:
The system performs preliminary actions by having cameras proactively adjust their video capture operations before objects enter their fields of view. Using graph map spatial relationships and event direction data, cameras predictively modify capture parameters in advance, ensuring objects are captured at high quality from the moment they enter view rather than waiting for trigger conditions.
Solution Approach 2:
The system implements feedback mechanisms where video capture update messages are sent between cameras based on detected video events. When one camera detects an object with determined event direction, it sends update messages to neighboring cameras via the controller, which then adjust their capture operations accordingly, creating a closed-loop feedback system that improves detection reliability.
2Quantity of substance
If video cameras operate at lower capture rates to reduce data usage, then bandwidth and storage are optimized, but early portions of video events and critical angles are missed
Solution Approach 1:
Cameras proactively adjust video capture operations before objects enter their fields of view by receiving video capture update messages that include child node identifiers from the graph map. This preliminary action ensures high-quality capture is already in progress when objects arrive, eliminating delays in capturing early event portions.
Solution Approach 2:
The system dynamically adjusts video capture parameters based on real-time events detected by neighboring cameras. Capture rates, field of view, and other parameters are modified dynamically in response to video events and their directional information, allowing the system to optimize between data volume and detection timing based on actual conditions.
3Measurement precision
If cameras modify field of view using PTZ capabilities upon object detection, then tracking accuracy improves, but objects that do not cross current field of view are missed
Solution Approach 1:
The controller acts as an intermediary that receives video events with event direction from one camera, queries the graph map for neighboring cameras with shared child nodes in that direction, and sends video capture update messages to those cameras. This intermediary coordination enables cameras to proactively adjust their fields of view toward predicted object locations, improving both tracking precision and coverage versatility.
Solution Approach 2:
Cameras proactively adjust their field of view and capture operations before objects enter their current fields of view by processing video capture update messages that contain child node identifier information from the graph map. This preliminary repositioning ensures objects are within the optimized field of view when they arrive, improving tracking precision without missing objects that would otherwise pass by.
Data Source
AI summary
Systems, video cameras, and methods for predictive adjustment of multi-camera surveillance video data capture based on graph maps are described. A plurality of networked video camera is deployed and represented in a graph map based on the video camera environment, with parent nodes corresponding to video cameras and child nodes corresponding to path intersections among the video cameras. When a video event is detected from video data for one of the video cameras, a video capture update message indicating a shared child node identifier is selectively sent to other video cameras to modify their video capture operations.