Radar Reflection Angle-Energy Processing for Object Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional radar data processing methods in automotive perception systems do not fully leverage the potential of radar data for improved object detection, classification, and semantic or instance segmentation, necessitating enhanced processing techniques.
Innovation Solution
The method involves determining the energy of radar reflections within a pre-determined angular region around a target angle, incorporating range and Doppler frequency information, and using neural networks to process these features, with data processing distributed across multiple hardware components to reduce communication load.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional radar data processing methods are used, then the system is simple and easy to implement, but object detection and classification performance cannot be improved
Solution Approach 1:
The patent segments the processing of radar data by distributing computations across multiple hardware components (satellite sensors and central perception unit). Different processing stages are divided between these components, with satellite sensors performing initial processing and the central unit performing higher-level analysis, thereby improving detection performance while managing system complexity through modular architecture
Solution Approach 2:
The patent introduces angular region energy determination as an additional dimension of feature extraction. By calculating energy within pre-determined angular regions around target angles and incorporating this into feature vectors, the system enriches the data dimensionality provided to neural networks, thereby improving classification and detection performance beyond conventional approaches
2Reliability
If more meaningful radar data features are provided, then object detection performance improves, but data transfer and processing load increases
Solution Approach 1:
The patent extracts only the most meaningful features from radar data by determining energy within specific angular regions around detected targets. This selective extraction approach provides enriched feature vectors to neural networks without transmitting all raw radar data, thereby improving classification accuracy while controlling data volume through targeted feature selection
Solution Approach 2:
The patent segments data processing across multiple hardware components to manage data volume. Satellite sensors perform initial processing and transmit processed results to a central perception unit, distributing the computational load and reducing the quantity of data that must be transferred across the entire system while still providing enriched features for accurate classification
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A Computer implemented method for processing radar reflections comprises the following steps carried out by computer hardware components: receiving radar reflections by at least one radar sensor; determining a target angle under which radar reflections related to a potential target are received by the at least one radar sensor; and determining an energy of radar reflections received by the at least one radar sensor under a pre-determined angular region around the target angle.