Signal Processing Device for Sensor Fusion Position Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In sensor fusion systems for automatic driving, there is a challenge in accurately determining the relationship between non-shared and shared detection objects across different sensors, such as cameras and millimeter-wave radars, due to differences in detection capabilities and measurement principles, leading to potential errors in determining the position of objects like vehicles and white lines.
Innovation Solution
A signal processing device that recognizes objects from images captured by cameras and radar devices, stores positional relationships, and compares these relationships to determine a final positional relationship between objects, using image analysis and radar analysis units to integrate information and correct for detection inaccuracies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sensor fusion is performed by combining detection results from camera and millimeter-wave radar, then distance accuracy is improved, but the relationship between non-shared detection objects (e.g., white line) and shared detection objects (e.g., vehicle) becomes erroneous
Solution Approach 1:
The patent segments the object detection process into separate handling for camera-only detected objects (white lines) and radar-only detected objects (vehicles), preserving their respective detection relationships independently before fusion, rather than attempting to fuse all objects uniformly which causes erroneous positional relationships
Solution Approach 2:
The patent introduces a transformation process that acts as an intermediary between the camera coordinate system and radar coordinate system, converting camera-detected object positions into radar coordinate system positions to enable accurate fusion while maintaining correct spatial relationships
2Adaptability or versatility
If detection objects are different among plurality of sensors, then each sensor can leverage its strengths (camera for direction, radar for distance), but it becomes difficult to correctly obtain positional relationships between non-shared and shared objects after fusion
Solution Approach 1:
The patent applies different detection and fusion strategies to different object types based on their detection characteristics: white lines (camera-only) maintain camera-based positional relationships while vehicles (radar-only) use radar-based relationships, with transformation applied only where needed to preserve local detection quality
Solution Approach 2:
The patent transforms the coordinate system from camera-based to radar-based dimension, enabling integration of detection results from different sensors while preserving accurate spatial relationships through coordinate transformation rather than direct fusion
Data Source
AI summary
Provided is a signal processing device with which, even when detection objects differ among a plurality of sensors, it is possible to correctly obtain the relationship between a non-shared detection object and a detection object that is shared after fusion.The present invention: stores a first positional relationship (for example, inside and outside a host vehicle lane) between a first object (for example, a vehicle ahead) recognized on the basis of an image captured by an imaging device 301, and a second object (for example, a white line); obtains the position of the first object on the basis of information about the first object recognized on the basis of information obtained from a sensor differing from the imaging device 301, and of information about the first object recognized on the basis of the image captured by the imaging device 301; obtains a second positional relationship (for example, inside and outside the host vehicle lane) between the obtained first object, and the second object recognized on the basis of the image captured by the imaging device 301; and compares the first positional relationship and the second positional relationship to determine a final positional relationship of the first object and the second object.