Heterogeneous Sensor Data Fusion via Joint Manifold Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current sensor fusion technologies face challenges in efficiently fusing high-dimensional, heterogeneous data streams from different sensor modalities, leading to information loss and reduced tracking accuracy due to the transformation from sensor data to decision-level fusion.
Innovation Solution
The method employs joint manifold learning to process data from multiple sensors, applying various manifold learning algorithms to reduce dimensions and extract intrinsic parameters, selecting the optimal algorithms for fusion, and applying them to fuse heterogeneous sensor data effectively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If decision-level fusion is used to combine outputs of multiple sensor modalities, then tracking accuracy is improved by incorporating decisions from different modalities, but information loss occurs during transformation from sensor data to decision
Solution Approach 1:
The patent transforms the fusion problem from decision space to feature space by embedding high-dimensional sensor data into a low-dimensional manifold representation. This dimensional transformation preserves essential feature information while achieving fusion, avoiding the information loss inherent in decision-level fusion.
Solution Approach 2:
The patent extracts intrinsic features from high-dimensional sensor data by applying manifold learning algorithms. This extraction process separates essential tracking information from redundant data, preserving salient features while reducing dimensionality and preventing information loss.
2Loss of information
If multiple manifold learning algorithms are applied to process joint manifold data, then information preservation is improved by discovering embedded low-dimensional features, but computational complexity increases due to processing multiple algorithms
Solution Approach 1:
The patent applies multiple manifold learning algorithms but selects only the optimal subset for final fusion based on performance evaluation. This partial application approach preserves essential information through multiple algorithmic perspectives while reducing computational complexity by not implementing all algorithms in the final system.
Solution Approach 2:
The patent performs preliminary evaluation of multiple manifold learning algorithms to identify optimal ones before applying them to actual sensor fusion. This preliminary action allows the system to discover which algorithms work best for specific sensor types and conditions, preserving information effectiveness while managing computational resources efficiently.
Data Source
AI summary
The present disclosure provides a method for joint manifold learning based heterogenous sensor data fusion, comprising: obtaining learning heterogeneous sensor data from a plurality sensors to form a joint manifold, wherein the plurality sensors include different types of sensors that detect different characteristics of targeting objects; performing, using a hardware processor, a plurality of manifold learning algorithms to process the joint manifold to obtain raw manifold learning results, wherein a dimension of the manifold learning results is less than a dimension of the joint manifold; processing the raw manifold learning results to obtain intrinsic parameters of the targeting objects; evaluating the multiple manifold learning algorithms based on the raw manifold learning results and the intrinsic parameters to determine one or more optimum manifold learning algorithms; and applying the one or more optimum manifold learning algorithms to fuse heterogeneous sensor data generated by the plurality sensors.