Multi-Sensor Data Fusion via Unified Coordinate Transformation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current multi-sensor data fusion methods, particularly in automotive perception systems, face challenges in achieving accurate fusion of data from sensors with different coordinate systems, leading to inconsistencies that reduce the effectiveness of neural network processing and result in loss of relevant information.
Innovation Solution
The method involves transforming sensor data into a unified coordinate system using a transformation rule that compensates for differences between individual sensor coordinate systems, allowing for the fusion of data in a common framework, thereby improving accuracy and enabling deeper neural network processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sensor data from multiple sensors with different coordinate systems are processed separately by neural networks, then processing accuracy for each sensor is maintained, but the fusion of sensor outputs loses relevant information and reduces overall accuracy
Solution Approach 1:
The patent applies preliminary coordinate transformation to convert sensor data from different coordinate systems into a unified coordinate system before neural network processing. This preliminary alignment ensures that when sensors are fused, their data is already in compatible spatial references, preventing information loss while maintaining individual processing accuracy through subsequent separate neural network streams.
Solution Approach 2:
The unified coordinate system acts as an intermediary framework that mediates between multiple sensor coordinate systems. By transforming all sensor data into this common reference frame, the patent enables accurate correspondence matching and fusion without direct coordination between disparate coordinate systems, thus preserving information while enabling effective integration.
2Measurement precision
If sensor data are transformed into a unified coordinate system before processing, then information consistency and fusion accuracy are improved, but processing complexity increases
Solution Approach 1:
The coordinate transformation is performed as a preliminary step before neural network processing, converting all sensor data to a unified coordinate system upfront. This eliminates the need for complex real-time coordinate transformations during fusion operations, reducing overall processing complexity while maintaining high fusion accuracy.
Solution Approach 2:
The processing pipeline is segmented into distinct stages: coordinate transformation, neural network processing, and fusion. By separating the coordinate system unification from the neural network processing, the patent manages complexity through modular design, where each stage handles a specific task independently.
3Measurement precision
If separate neural networks process each sensor output, then individual sensor processing accuracy is maintained, but robustness of perception results does not reach satisfactory level
Solution Approach 1:
The unified coordinate system serves as an intermediary that enables reliable correspondence between objects detected by different sensors. By providing a common spatial reference frame, it allows the system to robustly match and correlate detections across sensors, significantly improving perception reliability while preserving individual sensor processing accuracy.
Data Source
AI summary
A method of multi-sensor data fusion includes determining a plurality of first data sets using a plurality of sensors, each of the first data sets being associated with a respective one of a plurality of sensor coordinate systems, and each of the sensor coordinate systems being defined in dependence of a respective one of a plurality of mounting positions for the sensors; transforming the first data sets into a plurality of second data sets using a transformation rule, each of the second data sets being associated with a unified coordinate system, the unified coordinate system being defined in dependence of at least one predetermined reference point; and determining at least one fused data set by fusing the second data sets.