Obstacle Detection via Radar-Camera Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current obstacle detection systems in intelligent vehicles face inaccuracies in height direction positioning using millimeter-wave radar and monocular cameras, especially when obstacles are not in the same plane as the vehicle, leading to unreliable ranging results.
Innovation Solution
A method that combines millimeter-wave radar and camera data to detect obstacles by projecting radar position points onto an image, using deep convolution networks to filter and determine obstacle presence, and calculating distances, thereby enhancing accuracy and overcoming the limitations of single-sensor systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Length of stationary object
If millimeter-wave radar is used for obstacle detection, then long-range sensing capability is improved, but height direction positioning accuracy deteriorates due to lack of discrimination power
Solution Approach 1:
The patent combines millimeter-wave radar and monocular camera into a fused sensing system. The radar provides long-range detection capability while the camera supplies height direction discrimination. By merging the detection results of both sensors, the system achieves both long sensing range and accurate height positioning, resolving the contradiction between radar's range advantage and its height discrimination deficiency.
2Device complexity
If monocular camera is used for obstacle ranging, then system complexity is reduced, but ranging accuracy deteriorates when obstacle and vehicle are not in the same plane
Solution Approach 1:
The patent fuses monocular camera with millimeter-wave radar to compensate for the camera's ranging limitations. When the obstacle is not in the same plane as the vehicle, the radar's accurate distance measurement compensates for the camera's projection error, while the camera continues to provide height direction information. This merging maintains relatively simple system structure while improving ranging accuracy for obstacles at different elevations.
3Length of stationary object
If conventional radar sensing is used, then long-range detection is improved, but false detection increases due to inability to distinguish height direction
Solution Approach 1:
The patent merges radar detection results with camera image analysis to eliminate false detections. The radar detects targets at long range but cannot distinguish whether reflections come from ground or aerial objects. The camera provides visual confirmation and height information, allowing the system to filter out false detections from viaducts, billboards, or other non-obstacle reflections, thereby improving detection reliability while maintaining long-range capability.
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
This approach provides more accurate obstacle detection in front of the vehicle by integrating radar and camera data, improving height direction positioning and ranging accuracy, and is simple to implement and apply widely.
Implementation Method 1
a radio wave is transmitted by a millimeter-wave radar, and then a return wave is received, and position data of a target is measured based on the time difference between the transmitting and receiving
Implementation Method 2
a deep convolution network is used to filter and determine the presence of obstacles based on the projected position points on the image
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
A method and apparatus for detecting an obstacle, an electronic device, and a storage medium. A specific implementation of the method includes: detecting, by a millimeter-wave radar, position points of candidate obstacles in front of a vehicle; detecting, by a camera, a left road boundary line and a right road boundary line of a road on which a vehicle is located; separating the position points of the candidate obstacles according to the left road boundary line and the right road boundary line of the road on which the vehicle is located, and extracting position points between the left road boundary line and the right road boundary line; projecting the position points between the left road boundary line and the right road boundary line onto an image; and detecting, based on projection points of the position points on the image, a target obstacle in front of the vehicle.