Property State Tracking With Object Versions Across Time
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional imagery-based methods struggle to reliably distinguish between changes in property data due to actual property modifications or measurement errors, leading to inaccurate data aggregation and inability to disambiguate visual changes in property states.
Innovation Solution
A method and system for property state tracking that uses neural networks trained with self-supervised learning to generate consistent object representations despite appearance changes, associating different measurements with a universal object identifier and determining relationships between representations to track property versions over time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If conventional imagery-based methods are used to extract property data from images sampled over time, then property measurements can be obtained from multiple sources, but the appearance changes of properties (shadows, vegetation, imaging platform differences) result in inaccurate data and inability to disambiguate visual changes
Solution Approach 1:
The patent segments the property identification process into distinct components: extracting visual features from images, comparing features across time points, and separately determining whether changes represent actual property modifications versus measurement variations. This segmentation allows independent optimization of each component to improve overall reliability while aggregating measurements from multiple sources
Solution Approach 2:
The patent introduces an intermediary comparison mechanism that acts as a mediator between multiple property measurements. This intermediary process analyzes visual features and determines whether discrepancies represent actual changes or measurement variations, enabling reliable data aggregation from multiple sources while maintaining accuracy
2Quantity of substance
If multiple image captures are used to aggregate property data, then more measurements are available for analysis, but it becomes difficult to determine whether detected visual changes are due to data disparity or actual property changes
Solution Approach 1:
The patent implements a dynamic comparison process that adapts to different types of changes. The system dynamically determines whether visual discrepancies represent actual property modifications or measurement variations by analyzing visual features across multiple captures, making the detection process flexible and context-aware rather than static
Solution Approach 2:
The patent changes the parameters used for comparison by extracting and analyzing specific visual features (such as geometric properties, texture patterns, and structural characteristics) that are invariant to common measurement variations like shadows and lighting conditions, thereby improving the ability to detect actual property changes
3Quantity of substance
If property measurements are aggregated from different sources and time points, then more comprehensive property data can be obtained, but previously-determined property data may become invalid due to structural changes in the property
Solution Approach 1:
The patent implements a feedback mechanism where the system continuously monitors property measurements over time, compares visual features, and provides feedback to determine whether structural changes have occurred. This feedback loop enables the system to identify when previously-determined data becomes invalid and trigger appropriate responses, maintaining data validity throughout the aggregation process
Data Source
AI summary
In variants, the method can include: determining a timeseries of measurements of a geographic region; determining a set of object representations from the timeseries of measurements; and determining a timeseries of object versions based on relationships between the object representations.