Geospatial Event Sensing via Model Graph Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current trajectory analysis methods fail to detect patterns involving interactions between individuals, lack interpretability, and focus on individual aspects, making them inadequate for identifying events that require multiple attributes, leading to prevalent manual annotation in various domains.
Innovation Solution
A computer-implemented method and system for sensing events in dynamic activities, involving data acquisition, event detection through a model graph, and overlay generation, which automatically detects patterns, classifies events, and represents them graphically in augmented videos, allowing for objective analysis and improved interpretability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If trajectory analysis methods focus on classifying individual trajectories, then individual trajectory classification is improved, but the ability to detect patterns involving interactions between individuals deteriorates
Solution Approach 1:
The patent segments the trajectory analysis into multiple independent aspects (intersections, mobility modes, individual preferences) while maintaining the ability to detect patterns across all aspects simultaneously. This allows the system to preserve individual trajectory classification accuracy while also detecting interaction patterns that involve multiple aspects together.
Solution Approach 2:
The patent creates a universal trajectory analysis system that can handle multiple types of patterns and interactions simultaneously. The system is designed to detect not only individual trajectory characteristics but also interaction patterns between individuals, making it multi-functional and adaptable to various analysis needs without sacrificing precision in any single aspect.
2Extent of automation
If black box machine learning methods are used, then automation of trajectory analysis is improved, but interpretability and user ability to apply domain knowledge deteriorates
Solution Approach 1:
The patent incorporates feedback mechanisms that allow users to review and understand the analysis process. The system provides explanations for its findings and allows users to adjust parameters and re-run analyses, creating a feedback loop that maintains automation while improving interpretability and user control over the process.
Solution Approach 2:
The patent segments the complex machine learning process into interpretable components and steps. By breaking down the analysis into distinct modules (intersection detection, mobility mode classification, preference analysis), the system maintains automation while making each component understandable and modifiable by users with domain knowledge.
3Measurement precision
If existing methods focus on individual aspects of trajectories, then specific aspect analysis is improved, but the ability to identify events requiring multiple attributes deteriorates
Solution Approach 1:
The patent merges multiple individual aspect analyses (intersections, mobility modes, preferences) into a unified framework that can detect events requiring multiple attributes. The system combines results from different aspects to identify complex events such as traffic bottlenecks or dangerous situations that involve multiple characteristics simultaneously, while preserving the precision of individual aspect analysis.
Solution Approach 2:
The patent creates a universal analysis framework that can handle both simple individual aspect analysis and complex multi-attribute event identification. The system is designed to be flexible and adaptable, allowing users to analyze single aspects in detail or combine multiple aspects to detect complex events, without sacrificing precision in any individual aspect.
Data Source
AI summary
A computer-implemented method including processing data describing a sporting event in order to: evaluate a collection of models, each receiving as input at least part of the processed data, or part of an output of other models, or a combination thereof; compare outputs of the models to at least one entry in an event library, each entry including data at least indicative of a type of gameplay event in terms of criteria defined over output values of at least part of the collection of models; and output an event record whenever the criteria of an entry in the event library are satisfied by the outputs of the models, the event record including data at least indicative of the type of gameplay event represented by the corresponding entry in the event library. Also, the method is part of a computing system and a non-transitory computer-readable medium.