In-Vehicle FMCW Radar Classification Under Occlusion and Low Light
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Camera-based solutions for in-vehicle object detection fail to effectively identify objects in situations with occlusions or lighting issues, such as when a child is under a blanket, posing a risk of heatstroke due to ineffective detection.
Innovation Solution
Utilizing frequency modulated continuous wave (FMCW) radar for in-vehicle sensing and classification, which includes transmitting and receiving radar signals, generating classification data through FFT and Doppler analysis, and using classifiers to determine the sensing state and generate appropriate responses.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If camera-based solutions are used for in-vehicle object detection, then the system can provide visual detection capability, but the detection reliability deteriorates in situations with occlusions or lighting issues
Solution Approach 1:
The patent replaces the optical camera-based detection system with a radar-based detection system. Radar uses electromagnetic wave reflection principles to detect objects, making it immune to occlusions by blankets or clothing and unaffected by lighting conditions inside the vehicle. This substitution fundamentally resolves the reliability issues camera systems face in these scenarios.
2Reliability
If FMCW radar is used for in-vehicle sensing, then the detection capability under occlusion and lighting issues is improved, but the device complexity increases
Solution Approach 1:
The FMCW radar system performs multiple functions using a single integrated platform: it detects object presence, determines object location through range data, classifies objects using Doppler features and machine learning models, and triggers appropriate system responses. This multi-functionality consolidates what would otherwise require separate camera, microphone, and processing systems into one unified radar-based solution.
3Measurement precision
If radar signals are processed through multiple FFT operations and classification algorithms, then the measurement precision of object classification is improved, but the processing time increases
Solution Approach 1:
The system performs preliminary signal processing operations in advance: range FFT converts raw radar signals to range data, Doppler FFT extracts velocity information, and machine learning models are pre-trained with extensive datasets. During actual operation, these pre-processed features are quickly classified, significantly reducing real-time processing requirements while maintaining high classification precision.
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
Enhances the ability to accurately detect and classify objects within vehicles, overcoming occlusions and lighting issues, thereby reducing the risk of heatstroke by providing timely alerts and system responses.
Implementation Method 1
transmitting, via a FMCW radar sensor, radar transmission signals. The method includes receiving, via the FMCW radar sensor, radar reflection signals based on the radar transmission signals
Implementation Method 2
generating first range data by performing a fast Fourier transform (FFT) on a first set of chirps of a first frame using the digital signals
Implementation Method 3
generating Doppler data by performing the FFT on at least the first range data and the second range data. The method includes extracting Doppler features from the Doppler data. The Doppler features include a velocity of the radar subject
Data Source
AI summary
Systems and methods relate to in-vehicle sensing and classification via frequency modulated continuous wave (FMCW) radar. Radar reflection signals are received based on the radar transmission signals, and include a plurality of chirps across a plurality of frames. Range-FFT data is generated by performing a fast Fourier transform (FFT) on each chirp of a particular frame. Doppler-FFT data is generated by performing the FFT on the range-FFT data. Point cloud data of a radar subject is generated using the Doppler-FFT data. Doppler features, including a velocity of the radar subject, are extracted from the Doppler-FFT data. Classification data is generated to indicate a sensing state inside a vehicle based on the point cloud data and the Doppler features. The classification data includes class data that classifies the radar subject. A system response is generated to provide an action concerning the sensing state inside the vehicle based on the classification data.