Trajectory Matching Using Quality Classifiers and Confidence Values
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems face challenges in accurately matching over-the-top (OTT) and measurement report (MR) trajectories due to differences in data sources, timestamps, and sampling rates, which affects user location tracking and various location-based services.
Innovation Solution
A system and method for iteratively matching spatiotemporal trajectories using a quality classifier to classify trajectories into high and low quality, adjusting confidence values through machine learning, and re-inputting trajectories below a threshold for improved matching, employing distance measures and filters to enhance accuracy and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional trajectory matching methods are used to match OTT and MR trajectories, then the matching process can be completed, but the accuracy is low due to differences in data sources, timestamps, and sampling rates
Solution Approach 1:
The patent segments the trajectory matching process into distinct quality categories (high quality and low quality trajectories) and applies different processing strategies to each segment. High quality trajectories are matched directly, while low quality trajectories undergo additional processing steps including confidence value adjustment and re-matching, thereby improving overall accuracy without uniformly increasing complexity across all trajectories.
Solution Approach 2:
The patent introduces dynamic confidence values that are adjusted iteratively based on trajectory quality assessments. The confidence values are not static but are modified through multiple passes, with low quality trajectories receiving adjusted confidence values and being re-matched in subsequent iterations, allowing the system to adaptively improve matching accuracy.
2Measurement precision
If iterative matching with confidence value adjustment is implemented, then accuracy improves by 10%, but processing time increases
Solution Approach 1:
The patent applies different processing intensities to different subsets of trajectories based on their quality. High quality trajectories are processed quickly with minimal intervention, while only low quality trajectories undergo the more time-consuming confidence adjustment and re-matching processes. This localized approach to quality-based processing improves accuracy for problematic cases without uniformly increasing processing time for all trajectories.
Solution Approach 2:
The patent implements partial iteration by performing confidence value adjustment and re-matching only on low quality trajectories rather than all trajectories. This partial action approach applies the computationally intensive iterative process only where needed, achieving accuracy improvements while minimizing the overall time penalty that would result from applying the same process universally.
3Measurement precision
If all trajectories are processed with high computational effort, then matching accuracy improves, but processing speed decreases by a factor of 1.3
Solution Approach 1:
The patent differentiates between high quality and low quality trajectories and applies corresponding processing efforts. High quality trajectories receive minimal processing and are matched directly, maintaining high processing speed. Low quality trajectories receive the full computational effort including confidence value adjustment and iterative re-matching, achieving accuracy improvements only where needed rather than applying uniform high computational effort to all trajectories.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A system for iteratively matching a plurality of spatiotemporal trajectories of a plurality of users according to a threshold is provided. The system comprises at least one processor, being configured for: receiving a plurality of first spatiotemporal trajectories and a plurality of second spatiotemporal trajectories each generated based on location data. Using a quality classifier for classifying the plurality of first spatiotemporal trajectories into first low quality trajectories and first high quality trajectories and the plurality of second spatiotemporal trajectories into second low quality trajectories and second high quality trajectories the processor identifies a plurality of high quality trajectories pairs and a plurality of low quality pairs; associating to each pairs with a confidence value indicative of a disparity; adjusting the confidence values and re-inputting into the quality classifier all spatiotemporal trajectories which are part of a low quality pair or a high quality trajectories pair whose adjusted confidence value is below the threshold.