Queue Monitoring Correction Model for Occluded Video
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing queue monitoring systems using computer vision struggle with accuracy in occlusion conditions, where video capture devices have a head-on view of queues, leading to incomplete data and difficulty in determining queue length and wait times without modifying the camera's view.
Innovation Solution
A system that employs a correction model to adjust for detection errors caused by occlusion, using object detection and tracking systems to determine queue size and wait times, allowing for accurate monitoring without reconfiguring the video capture devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a head-on view is used by video capture devices, then queue monitoring is simplified, but detection accuracy deteriorates due to occlusion conditions
Solution Approach 1:
A correction model is introduced as an intermediary between the object detection system and the final queue monitoring results. The correction model receives detection data from the head-on view camera and applies corrections based on occlusion patterns, thereby maintaining the simplicity of the head-on view setup while improving detection accuracy through computational correction
Solution Approach 2:
The system changes parameters related to detection confidence thresholds and correction factors based on detected occlusion conditions. When occlusion is detected, the system adjusts detection parameters and applies correction models that modify the interpretation of detected objects, allowing accurate queue length estimation despite the fixed head-on camera view
2Measurement precision
If the video capture device view is modified to reduce occlusion, then detection accuracy improves, but device complexity and reconfiguration effort increase
Solution Approach 1:
Instead of modifying the physical camera setup, a software-based correction model serves as an intermediary that processes the existing head-on view data. This approach achieves improved detection accuracy through algorithmic correction rather than physical reconfiguration, thereby avoiding increased device complexity
Solution Approach 2:
The patent replaces the mechanical approach of physically repositioning or reconfiguring cameras with a computational approach. The correction model uses image processing and detection algorithms to compensate for occlusion effects, substituting mechanical system changes with software-based solutions that maintain the original camera installation
3Loss of information
If the video capture device view is modified to include the entire queue, then measurement completeness improves, but ease of operation deteriorates due to frustration of other uses
Solution Approach 1:
The system segments the video feed processing into different functional streams. The head-on view camera continues to provide video feed for general monitoring purposes, while a separate object detection and tracking system extracts queue-specific information. This segmentation allows the same video feed to serve multiple purposes without compromising usability for either function
Solution Approach 2:
An object detection and tracking system serves as an intermediary that extracts queue monitoring data from the existing video feed without altering the camera view. This intermediary layer processes the video data to identify and track queue members, providing complete queue state information while leaving the original video feed intact for other uses
Data Source
AI summary
A system may be configured to perform enhanced detection of occluded objects in a multiple object detection system. In some aspects, the system may determine a first size of a queue captured in a video frame based on an object detection system and a video capture device, and determine a second size of the queue based on the first size of the queue and a correction model. Further, the system may detect a person at a front of the queue based on an object tracking system, determine a first period of time spent by the person at the front of the queue based on an object tracking system, and estimate a second period of time spent in the queue based at least in part on the first period of time and the second size of the queue.