Neural Network Object Detection Using Radar Spectra and Occupancy Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object detection systems using neural networks and radar data suffer from inadequate performance due to reliance on preset parameters and lack of robustness in varying conditions, particularly in image processing and radar data fusion.
Innovation Solution
A method for training a neural network using occupancy maps and mixed radar spectra, generated from combined camera and radar data, to enhance object detection by incorporating geometric dimensions, azimuth angles, elevation angles, and radial velocities, thereby improving classification accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional radar algorithms (CFAR) with preset parameters are used for object detection, then the system is simple to operate, but the detection accuracy and reliability deteriorate under varying conditions
Solution Approach 1:
The patent introduces neural networks as an intermediary between radar data and object detection decisions. The neural network is trained offline using labeled radar data to learn optimal detection patterns, then serves as a mediator that processes new radar data without requiring manual parameter adjustment. This resolves the contradiction by maintaining operational simplicity while significantly improving detection reliability through learned patterns rather than rigid preset parameters.
Solution Approach 2:
The patent applies preliminary action by training the neural network offline before deployment using extensive labeled radar data. This pre-training phase allows the system to learn optimal detection patterns in advance, so that during actual operation, the pre-trained network can reliably detect objects without requiring real-time parameter tuning or complex adjustments, thus maintaining simplicity while improving reliability.
2Measurement precision
If neural networks are trained with sufficient mixed spectra containing ground truth information, then object detection accuracy improves, but the training data generation complexity increases
Solution Approach 1:
The patent applies self-service by having the system generate its own training data automatically. The neural network is trained using radar data that has been automatically labeled with ground truth information about object positions and characteristics. This automated self-labeling process eliminates the need for manual annotation of training data, reducing training complexity while maintaining high detection accuracy through sufficient and relevant training examples.
Solution Approach 2:
The patent uses partial action by focusing the training process on the most critical features and data elements needed for detection accuracy. Rather than processing all possible radar data parameters equally, the system identifies and prioritizes the key spectral features and ground truth information that most directly impact detection performance, thereby achieving high accuracy without overwhelming training complexity.
3Adaptability or versatility
If multiple dimensions of radar data (distance, azimuth angle, elevation angle, radial velocity) are integrated into training data, then classification accuracy improves, but the data processing complexity increases
Solution Approach 1:
The patent merges multiple dimensions of radar data (distance, azimuth angle, elevation angle, radial velocity) into a unified mixed spectrum representation that serves as comprehensive training input for the neural network. By combining these different data dimensions into a single integrated feature space, the system achieves versatile multi-dimensional classification capability without requiring separate processing pipelines for each parameter, thus managing complexity through unification rather than multiplication of processing steps.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The proposed method enhances the robustness and accuracy of object detection by leveraging multiple dimensions of radar data, enabling effective classification and localization of objects independent of image conditions.
Implementation Method 1
The radial velocity is determined via a frequency shift between the transmitted radar signal and the reflected radar signal. The frequency shift in question results from the Doppler effect of moving objects.
Data Source
AI summary
In a method for training a neural network for detecting an object, geometric dimensions of a test object from an object class are captured, and during a time period, recordings of the test object are generated by a plurality of cameras. From the captured geometric dimensions and the generated recordings, occupancy maps are generated. By a radar device, a radar signal is transmitted, and a radar signal reflected by the test object is received. The transmitted radar signal and the received radar signal are mixed into a complex baseband to form a mixed signal. A complex four-dimensional mixed spectrum of the mixed signal is calculated. From the complex four-dimensional mixed spectrum, a first complex two-dimensional partial spectrum and a second complex two-dimensional partial spectrum are calculated. The occupancy maps and the partial spectra are fusioned to form training data. The training data are fed to the neural network.