Millimeter-Wave Radar Target Feature Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing radar technologies face limitations in road surface target recognition using millimeter-wave radar, particularly in extracting features from point cloud data and Range Doppler (RD) maps, which are limited by hardware constraints and struggle to distinguish targets similar in physical shape and motion state.
Innovation Solution
A method that fully extracts features from raw radar echo information using an improved cell-averaging CFAR algorithm, DBSCAN clustering, centroid condensation, and Kalman filtering to generate richer classification features by processing raw radar echoes, removing ground clutter, detecting targets, and tracking them across multiple frames, incorporating time sequence information from two successive RD maps.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If point cloud data processing methods are used for target recognition, then target classification can be performed, but the acquisition process is complex and requires high hardware conditions including high angular resolution of radar
Solution Approach 1:
The patent extracts and utilizes raw radar echo information directly from the radar signal processing chain, bypassing the need for complex point cloud generation and high angular resolution hardware. By working with the raw echo data and RD map information that is already available from standard radar hardware, the method eliminates the requirement for additional high-resolution angular sensors while maintaining target classification capability
Solution Approach 2:
The patent makes the feature extraction method universally applicable to standard millimeter-wave radar systems without requiring specialized high-resolution angular resolution hardware. The method works with conventional radar configurations by utilizing the RD map and raw echo information that are produced by standard radar processing pipelines
2Device complexity
If feature extraction is performed using a single RD map, then processing is simplified, but effective features cannot be obtained from objects that are similar in motion state and physical structure
Solution Approach 1:
The patent transitions from single-frame RD map analysis to multi-frame temporal sequence analysis. By incorporating the time dimension and comparing successive RD maps, the method extracts dynamic features that capture changes in target motion state and physical structure over time, enabling discrimination between similar targets that would be indistinguishable in a single static frame
Solution Approach 2:
The patent performs preliminary tracking and candidate area selection across multiple frames before final feature extraction. By pre-identifying target regions and their temporal evolution through tracking algorithms, the method prepares enhanced feature sets that capture discriminative characteristics of similar targets before the classification decision is made
3Loss of information
If raw radar echo information is fully utilized for feature extraction, then richer classification features are obtained, but processing complexity increases
Solution Approach 1:
The patent segments the raw radar echo information processing into distinct functional modules: RD map generation, ground clutter removal using CLEAN algorithm, target detection using CFAR, tracking across frames, and feature extraction from candidate areas. This modular segmentation manages processing complexity by organizing the information-rich processing pipeline into manageable, independent stages that can be implemented and optimized separately
Data Source
AI summary
The present disclosures discloses a method of target feature extraction based on millimeter-wave radar echo, which mainly solves the problems that techniques in the prior art cannot fully utilize raw radar echo information to obtain more separable features and cannot accurately distinguish targets with similar physical shapes and motion states. The method is implemented as follows: acquiring measured data of targets, generating an original RD map, and removing ground clutter of the map; sequentially performing target detection, clustering and centroid condensation on the RD map after the ground clutter removal; acquiring a continuous multi-frame RD maps and carrying out the target tracking; according to the tracking trajectory, selecting candidate areas and extracting features based on a single piece of RD map and features based on two successive RD maps, respectively.