Object Identifier Determination Using Sensor Context
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for re-identifying objects between multiple sensors in military and driver assistance systems rely on manual operator selection, which is slow and potentially unreliable.
Innovation Solution
A computer-implemented method that detects objects, obtains sensor signals containing positional and contextual information, and uses a machine learning module to determine a unique identifier or fingerprint associated with the object, improving accuracy and reliability of object identification and tracking.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual operator selection is used for object re-identification, then the system can identify objects between multiple sensors, but the process is slow and potentially unreliable
Solution Approach 1:
The patent replaces the manual mechanical process of operator selection with an automated machine learning system. The ML module processes sensor signals containing positional and contextual information to automatically generate object identifiers, eliminating the need for human operators to manually match objects across sensors. This substitution dramatically reduces re-identification time while maintaining or improving reliability through consistent automated processing.
Solution Approach 2:
The system enables self-service by allowing objects to be automatically identified and tracked through their own unique characteristics. The machine learning module extracts features directly from sensor data about each object (position, context, appearance) and generates identifiers without external human intervention. The system serves itself by autonomously performing the re-identification task that previously required human operators.
2Measurement precision
If only positional information is used for object identification, then the system is simple to implement, but accuracy decreases in crowded areas with numerous objects
Solution Approach 1:
The patent merges multiple types of information (positional information, contextual information, and visual features) into a comprehensive object identification system. The machine learning module processes and integrates these different data streams to generate robust object identifiers. This combination allows the system to accurately distinguish between numerous objects in crowded areas by considering their spatial relationships, contextual attributes, and visual characteristics simultaneously.
Solution Approach 2:
The machine learning module serves multiple functions: it processes positional data, analyzes contextual information, extracts visual features, and generates object identifiers. This multi-functional approach allows a single system component to handle diverse information types and perform comprehensive object identification, improving accuracy without proportionally increasing system complexity.
3Reliability
If a single type of information is used for object identification, then the system is simpler, but reliability decreases when sensor information is temporarily lost or occluded
Solution Approach 1:
The system implements beforehand cushioning by maintaining multiple types of object information (positional, contextual, visual features) simultaneously. When one information source becomes unavailable due to occlusion or sensor failure, the system has redundant information available to continue tracking the object. This preparatory accumulation of diverse data types cushions against information loss and maintains reliable object identification.
Solution Approach 2:
The machine learning module dynamically adjusts which information parameters are used for identification based on availability. When sensor information is temporarily lost, the system can switch to relying more heavily on other available parameters (e.g., using contextual information when visual features are occluded, or using positional data when appearance information is unavailable). This flexibility maintains identification reliability despite changes in information availability.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
According to an embodiment of the present disclosure, there is provided a computer-implemented method of determining an object identifier, the method comprising: detecting an object; obtaining one or more sensor signals, wherein the sensor signals represent data including at least: positional information related to the object, and contextual information relating to the object; and providing the data to a machine learning module to determine an identifier, in particular a unique identifier or fingerprint associated with the object based on the data.