Location-aware event detection using overlapping ROIs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Video analytics surveillance systems face challenges in detecting repetitive events at specific locations, such as checkout processes, due to the un-oriented interaction between the cashier's hands and the belt area, making it difficult to define an appropriate region of interest (ROI) for event models, as overly large ROIs include irrelevant features and overly small ones miss relevant events.
Innovation Solution
The use of multiple overlapping regions of interest (ROIs) and multiple-instance learning techniques to cover all possible locations for events, allowing for the construction of location-aware event models that can detect events and determine the associated regions, thereby addressing the challenge of defining an appropriate ROI.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If an overly large ROI is used to cover all possible event locations, then event detection coverage is improved, but the number of irrelevant features increases
Solution Approach 1:
The patent divides the transaction area into multiple overlapping ROIs (e.g., first ROI, second ROI, third ROI) instead of using a single large ROI. Each ROI is assigned to specific event types (pickup, scan, drop) and spatial locations, allowing the system to process only relevant features for each region while maintaining comprehensive event detection coverage across the entire checkout area.
2Object-generated harmful factors
If an overly small ROI is used to reduce irrelevant features, then feature relevance is improved, but event detection completeness deteriorates
Solution Approach 1:
The transaction area is segmented into multiple overlapping ROIs that collectively cover all possible event locations. The first ROI covers pickup areas, the second ROI covers scan areas, and the third ROI covers drop areas, with overlapping regions ensuring complete coverage. This segmentation allows the system to maintain small, relevant ROIs for each event type while ensuring no events are missed.
Solution Approach 2:
The patent introduces a spatial dimension by creating multiple ROIs at different locations and orientations within the transaction area. Instead of using a single large ROI, the system creates a distributed set of smaller ROIs that together provide comprehensive coverage, effectively adding spatial diversity to the detection approach.
3Adaptability or versatility
If a sliding window approach is used to exhaustively test every possible location, then event detection completeness is improved, but computational efficiency deteriorates
Solution Approach 1:
The patent performs preliminary action by pre-defining a set of overlapping ROIs based on the checkout area layout and event types before actual event detection. These ROIs are strategically positioned and sized to cover all possible event locations in advance, eliminating the need for exhaustive sliding window testing during runtime and significantly improving computational efficiency.
Data Source
AI summary
Techniques for detecting one or more events are provided. The techniques include using multiple overlapping regions of interest on a video sequence to cover a location for one or more events, wherein each event is associated with at least one of the multiple overlapping regions of interest, applying multiple-instance learning to the video sequence to select one or more of the multiple overlapping regions of interest to construct one or more location-aware event models, and applying the models to the video sequence to detect the one or more events and to determine the one or more regions of interest that are associated with the one or more events.