Radar-Camera Fusion for Stable AI Object Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automotive radars face challenges in accurately classifying objects due to low horizontal resolution and difficulty in annotation, while cameras struggle with range accuracy and performance in adverse weather conditions, limiting effective object detection and classification.
Innovation Solution
A radar apparatus that generates fusion data by combining radar and camera data, using an artificial intelligence module trained on this fusion data for improved object classification, enabling stable detection and classification even in bad weather or at night, and allowing for camera-independent object classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If automotive radar is used for object detection, then range accuracy and stability in adverse weather are improved, but horizontal resolution and object classification capability deteriorate
Solution Approach 1:
The patent combines radar and camera data through data fusion to create comprehensive object detection. The radar provides stable range and velocity information while the camera supplies high-resolution visual data for object classification, resolving the contradiction between detection stability and classification capability.
Solution Approach 2:
The system integrates multiple sensing functions into a unified detection framework. The radar apparatus not only detects objects with high reliability but also classifies them by fusing camera image data, making the system universally capable of both detection and classification tasks.
2Measurement precision
If automotive camera is used for object classification, then horizontal resolution and classification capability are improved, but range accuracy and performance in adverse weather deteriorate
Solution Approach 1:
The system merges camera data with radar data to compensate for the camera's weaknesses. When weather conditions deteriorate, the radar's stable detection data supplements the camera's visual information, maintaining reliable object detection and classification across all conditions.
Solution Approach 2:
The radar data acts as an intermediary that bridges the gap between visual detection and reliable detection. By fusing radar range and velocity information with camera image data, the system achieves both high resolution and weather-resistant performance.
3Reliability
If radar data alone is used for object classification, then detection stability is maintained, but classification accuracy deteriorates due to insufficient information
Solution Approach 1:
The system combines radar detection data with camera image data to create fusion data that contains both stable detection information and detailed object characteristics. This merging eliminates information loss while maintaining detection stability.
4Measurement precision
If fusion data generation is implemented, then object classification accuracy is improved, but system complexity increases
Solution Approach 1:
The system achieves multi-functional capability through a unified data fusion framework that handles both detection and classification tasks, reducing overall system complexity despite the sophisticated processing required.
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 solution enhances object detection and classification performance by leveraging the strengths of both radar and camera systems, providing stable performance in adverse conditions and improving the accuracy of object classification using AI-trained modules.
Implementation Method 1
a transceiver that transmits a radar signal to an outside of the vehicle and receives a radar signal reflected from an object
Data Source
AI summary
A radar apparatus includes a transceiver that transmits a radar signal and receives a radar signal reflected from an object; a signal processing unit that processes the reflected radar signal to detect the object; a fusion data generation unit that generates fusion data based on radar data and camera data; and a classification unit that classifies the detected object using an artificial intelligence module trained based on the generated fusion data.