Sensor Data Fusion via Element-Wise Addition for ADAS
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for fusing sensor data in ADAS/AD systems, particularly using artificial neural networks, face challenges in efficiently processing input data from different acquisition ranges and resolutions, leading to suboptimal object recognition and feature extraction.
Innovation Solution
A method that involves receiving input sensor data from overlapping acquisition ranges, determining feature maps for each representation, computing output feature maps through convolution, and fusing these maps element-by-element while considering the positional overlap, to produce a unified fused feature map for subsequent processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If sensor data from multiple acquisition ranges and resolutions are processed separately, then processing complexity is reduced, but object recognition accuracy deteriorates because objects in overlapping regions cannot be properly identified as belonging to the same object
Solution Approach 1:
The patent merges sensor data from multiple acquisition ranges and resolutions into a single unified representation before processing. By combining the first sensor data including a first region and the second sensor data including a second region into one integrated dataset, the system maintains object continuity across different regions while enabling comprehensive feature extraction, thus resolving the contradiction between processing simplicity and recognition accuracy
Solution Approach 2:
The patent introduces a new dimension of integration by creating a unified feature space that encompasses multiple acquisition ranges and resolutions. This dimensional expansion allows the neural network to process all sensor data together in a coherent framework, enabling accurate object identification across overlapping regions without requiring separate processing pipelines
2Measurement precision
If high resolution is used for all sensor data, then object identification accuracy is improved, but computing power requirements increase significantly
Solution Approach 1:
The patent applies local quality by processing different regions of the sensor data at appropriate resolution levels. By integrating sensor data from multiple acquisition ranges where each region can be processed at its optimal resolution, the system achieves high object identification accuracy in critical areas while reducing overall computing power requirements through selective resolution processing
Solution Approach 2:
The patent segments the sensor data processing into multiple components corresponding to different acquisition ranges and resolutions. By dividing the overall processing task into manageable segments that can be independently processed and then integrated, the system maintains high accuracy for important features while reducing the total computational burden through efficient resource allocation across segments
Data Source
AI summary
A method and system for fusing data from at least one sensor, including: receiving input sensor data, wherein the input sensor data include: first and second representations including first and second regions, respectively, of a scene, wherein the first and second regions overlap one another but are not identical; determining first and second feature maps on the basis of the first and second representations, respectively; computing first and second output feature maps by a convolution of the first and second feature maps, respectively; and computing a fused feature map through element-by-element addition of the first and second output feature maps, wherein the relative position of the first and second regions to one another is used, such that the elements in the region of overlap are added; and outputting the fused feature map. The method is runtime-efficient and deployed to fuse data from environment sensors for a vehicle's ADAS/AD system.