Radar Resolution Increase Model Using Neural Network
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current radar systems face limitations in resolution due to restricted bandwidth and the number of radar sensors, leading to reduced accuracy in object detection and recognition, especially in dimensions like Doppler velocity, range, and angular resolution.
Innovation Solution
A processor-implemented method is developed to enhance radar resolution by training a radar resolution increase model using a neural network, which generates high-resolution output data from low-resolution input through preprocessing techniques like FFT and DBF, and incorporates direction-of-arrival information for improved subspace type-based estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If radar bandwidth and number of radar sensors are restricted, then device complexity and cost are reduced, but measurement precision and resolution deteriorate
Solution Approach 1:
A neural network model is introduced as an intermediary between low-resolution radar input data and high-resolution radar output data. The model learns the mapping relationship through training with paired datasets, enabling resolution enhancement without adding physical radar sensors or increasing bandwidth.
Solution Approach 2:
The patent replaces the traditional mechanical approach of improving radar resolution by increasing bandwidth and sensor count with a computational approach using deep learning. The neural network model substitutes for additional hardware resources, achieving high-resolution output through software-based processing of low-resolution input.
2Measurement precision
If more radar sensors and bandwidth are used, then radar resolution is improved, but device complexity and cost increase
Solution Approach 1:
The radar data processing is segmented into two distinct stages: (1) data collection using a limited number of radar sensors with restricted bandwidth, and (2) resolution enhancement using a trained neural network model. This segmentation allows the system to achieve high-resolution output without requiring high-resolution input from multiple sensors.
Solution Approach 2:
The neural network model serves as an intermediary that transforms low-resolution radar data into high-resolution radar data. By learning the complex mapping relationship from training datasets, the model enables resolution enhancement without adding physical radar sensors.
3Measurement precision
If high-resolution radar data is collected directly, then measurement precision is improved, but loss of time and processing complexity increase
Solution Approach 1:
The neural network model is trained in advance using paired datasets of low-resolution and high-resolution radar data. This preliminary training phase enables the model to learn the resolution enhancement mapping, so that during actual operation, only forward propagation through the trained model is needed, significantly reducing real-time processing time compared to iterative optimization methods.
Data Source
AI summary
A method of increasing a resolution of radar data is provided. The method of training a radar resolution increase model comprises generating a high-resolution training ground truth and a low-resolution training input from original raw radar data based on information corresponding to at least one of dimensions defining the original raw radar data, and training the resolution increase model based on the high-resolution training ground truth and the low-resolution training input. A radar data processing device generates high-resolution output data from low-resolution input data based on a trained resolution increase model.