Vehicle Object Distance Estimation by Camera-Radar Overlap
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle systems face challenges in accurately fusing sensor data from different types of surround sensors to precisely recognize and track surroundings objects, as image data from cameras provide detailed attributes but imprecise distance, while radar and lidar data offer precise distance but limited attributes, leading to inconsistent object detection and tracking.
Innovation Solution
A processing unit is implemented to synchronize and transform image and sensor data into a common coordinate system, allowing the camera object to be displaced and scaled to assess overlap with sensor data points, thereby determining accurate object distance and enhancing detection quality by creating an occupancy grid that indicates object presence and type.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If image data from a camera is used to detect surroundings objects, then detailed attributes such as object type, color, and shape can be ascertained, but the distance of the object from the vehicle cannot be determined precisely
Solution Approach 1:
The patent combines image data from a camera with sensor data from a surround sensor (radar or lidar) to create a fused data representation. The image data provides detailed object attributes while the sensor data provides precise distance information. By merging these complementary data sources in a common coordinate system, the system achieves both rich attribute information and accurate distance measurement simultaneously.
2Measurement precision
If sensor data from a radar or lidar sensor is used to detect surroundings objects, then the distance of the object from the vehicle can be ascertained precisely, but only relatively few attributes of the object can be ascertained
Solution Approach 1:
The patent merges sensor data from radar or lidar (providing precise distance) with image data from a camera (providing detailed attributes). The processing unit combines these data sources in a common coordinate system, allowing the system to achieve both accurate distance measurement and comprehensive object attribute information simultaneously.
3Loss of information
If image data and sensor data from different sensors are fused, then comprehensive object information can be obtained, but sensor calibration errors and temporal synchronization issues affect detection reliability
Solution Approach 1:
The patent dynamically adjusts the object position within an angle range rather than using a fixed position. The processing unit determines an angle range based on the image data and then evaluates sensor data at multiple positions within this range. This dynamic approach compensates for calibration errors and synchronization issues by finding the best match among multiple possible positions.
Solution Approach 2:
The system uses feedback by evaluating the overlap between the object representation and sensor data points at different positions within the angle range. The processing unit determines the object distance based on which position provides the best overlap match, creating a feedback mechanism that refines the detection and compensates for errors in the sensor calibration or timing.
Data Source
AI summary
A controller for a vehicle is configured to determine image data from a camera of the vehicle which indicates an environment of the vehicle starting from a reference point. The controller further detects at least one camera object in the environment of the vehicle based on the image data. The controller is also configured to determine sensor data from a distance-sensing environment sensor of the vehicle. The camera object can be shifted to a plurality of different distances from the reference point. A corresponding plurality of values of a degree of overlap between the camera object and the sensor data can be determined for the plurality of different distances. The controller is also configured to determine an object distance between the camera object and the reference point based on the plurality of values of the degree of overlap.