Steered Multi-Sensor Perception for Long-Range Object Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in accurately detecting and identifying objects at longer distances due to reduced resolution of images captured by sensors, which can lead to inefficient and potentially unsafe control systems.
Innovation Solution
The use of multiple imaging and ranging sensors with different field of views and resolutions, where a second sensor with a narrower field of view and higher resolution can be directed to capture specific objects of interest, improving identification accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If a single sensor with wide field of view is used to capture the scene, then the coverage area is maximized, but the resolution and identification accuracy of distant objects deteriorates
Solution Approach 1:
The system divides the sensing task into two segments: a first sensor (camera) with wide field of view for general scene coverage, and a second sensor (LIDAR) with narrow field of view for high-precision measurement of specific objects. This segmentation allows each sensor to optimize for its specific function, resolving the contradiction between coverage area and measurement precision.
Solution Approach 2:
The system transitions from two-dimensional image data from a single camera to three-dimensional point cloud data from LIDAR. This dimensional change enables precise depth measurement and object identification at longer distances, overcoming the resolution limitations of wide-field imaging sensors.
2Measurement precision
If a sensor with narrow field of view and high resolution is used, then the object identification accuracy is improved, but the detection range and coverage area is reduced
Solution Approach 1:
The first sensor (camera) performs preliminary scanning of the entire scene to identify objects of interest. Based on this preliminary detection, the second sensor (LIDAR) is then directed to specific regions for detailed measurement. This preliminary action allows the high-precision sensor to focus only on relevant areas, maintaining both accuracy and coverage.
Solution Approach 2:
The LIDAR sensor is mounted on a movable platform that can dynamically adjust its field of view direction based on object locations detected by the camera. This dynamic repositioning allows the narrow-field sensor to cover different areas sequentially, effectively increasing the total detection coverage while maintaining high resolution.
3Productivity
If the vehicle speed is increased for efficient transportation, then the productivity is improved, but the time available for object detection and response is reduced
Solution Approach 1:
The wide-field camera continuously scans the scene ahead to preliminarily identify potential obstacles and road conditions at long distances. This early detection provides advance warning, allowing the control system to prepare appropriate responses well before the vehicle reaches the objects, thus maintaining safety even at high speeds.
Solution Approach 2:
The system uses the camera as an intermediary sensor that detects objects at long range and triggers the LIDAR for detailed measurement only when necessary. This intermediary approach reduces the overall detection time by avoiding continuous high-precision scanning of all areas, enabling faster response times that support higher vehicle speeds.
Data Source
AI summary
Disclosed are methods, systems, and non-transitory computer readable media that control an autonomous vehicle via at least two sensors. One aspect includes capturing an image of a scene ahead of the vehicle with a first sensor, identifying an object in the scene at a confidence level based on the image, determining the confidence level of the identifying is below a threshold, in response to the confidence level being below the threshold, directing a second sensor having a field of view smaller than the first sensor to generate a second image including a location of the identified object, further identifying the object in the scene based on the second image, controlling the vehicle based on the further identification of the object.