Multi-state target tracking for video surveillance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional intelligent video surveillance systems face challenges in tracking targets that are occluded or stationary, as they often lose track or incorrectly identify targets due to occlusions and assume stationary targets will not move again, leading to inefficiencies and inaccuracies.
Innovation Solution
A multi-state target tracker that considers targets to be in one of several states (visible, occluded, disappeared, etc.) and uses specific modules to handle transitions between these states, enabling reliable tracking through multiple frames and user-defined rules for event detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional target tracking methods are used, then the system can track moving targets, but the tracking reliability deteriorates when targets are occluded or stationary
Solution Approach 1:
The patent applies dynamics by transitioning from static tracking assumptions to dynamic state management. The system continuously updates target states (visible, occluded, disappeared, stationary) based on real-time detection, allowing the tracker to adapt its behavior to current conditions. This dynamic state transitions framework enables reliable tracking through occlusions by predicting target states rather than assuming constant visibility.
Solution Approach 2:
The patent changes the parameter of target state classification from binary (tracked/not tracked) to multi-state (visible, occluded, disappeared, stationary). This parameter transformation allows the system to differentiate between various target conditions and apply appropriate handling strategies for each state, improving reliability for occluded and stationary targets.
2Reliability
If prediction is used to handle occlusion, then short-term occlusion can be managed, but the method fails when occlusion duration is long or velocity changes
Solution Approach 1:
The patent applies preliminary action by pre-defining multiple target states and transition rules before occlusion occurs. When a target becomes occluded, the system has already established prediction mechanisms for that state, allowing it to handle occlusions of any duration without relying solely on short-term velocity prediction. The state framework prepares the system in advance for various occlusion scenarios.
Solution Approach 2:
The system dynamically adjusts tracking strategies based on the current state and duration of occlusion. For short occlusions, prediction is used; for long occlusions, the system maintains the occluded state with updated timing information. This dynamic adaptation allows the system to tolerate varying occlusion durations without losing track of targets.
3Reliability
If stationary targets are burned into background, then false tracking is reduced, but the system loses the ability to detect when targets start moving again
Solution Approach 1:
The patent makes the stationary target handling dynamic by continuously monitoring targets in the stationary state and detecting transitions to moving state. Rather than permanently burning targets into background, the system maintains them in a reversible stationary state, allowing automatic detection when they begin moving again. This dynamic approach eliminates false tracking while preserving movement detection capability.
Solution Approach 2:
The system implements feedback by continuously checking the state of stationary targets and detecting when they transition to moving state. This feedback mechanism allows the system to respond to target movement by exiting the stationary state and resuming normal tracking, preventing false tracking while maintaining adaptability to target behavior changes.
4Measurement precision
If conventional tracking is used, then simple moving targets are tracked accurately, but performance deteriorates in complex scenarios with occlusions and stationary targets
Solution Approach 1:
The patent applies segmentation by dividing the tracking problem into distinct state handling modules. Each target state (visible, occluded, disappeared, stationary) has its own dedicated processing logic and transition rules. This segmentation allows the system to maintain high measurement precision for simple targets while systematically handling complex scenarios through modular state-based processing.
Solution Approach 2:
The system dynamically selects appropriate tracking strategies based on the current target state. For visible targets, conventional accurate tracking is used; for occluded targets, prediction and state maintenance are applied; for stationary targets, special monitoring is activated. This dynamic strategy selection maintains measurement precision across different scenarios while managing complexity through state-based organization.
Data Source
AI summary
A method of video analysis may comprise tracking a state of each target in a video through multiple frames of said video, each state indicating a visibility condition of a target.