Object Distance Determination via Sensor Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for autonomously driving vehicles lack efficient and reliable systems for determining the distance of objects around the vehicle, particularly in real-time, as they fail to accurately convert 2D camera image information into 3D positional data relative to the ego vehicle.
Innovation Solution
A computer-implemented method using machine learning models, such as SSD or YOLO neural networks, to classify objects and combine this classification with distance sensor data for accurate distance determination, employing a hash table for correction and reprojection onto a Bird's Eye View coordinate system, allowing for real-time detection and localization of objects in a 3D space.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If 2D camera image information is used to determine object distance, then the system can operate in real-time with low cost, but the measurement precision of distance is insufficient
Solution Approach 1:
The patent combines 2D camera images with 3D point cloud data from depth sensors (LiDAR, stereo cameras, or time-of-flight sensors) to create a fused representation that preserves both the real-time processing capability of 2D images and the accurate distance information from 3D sensors. This merging allows the system to maintain high detection speed while achieving precise distance measurement.
Solution Approach 2:
The patent introduces intermediate processing steps including reprojection of 2D detected objects onto 3D space, assignment of 3D coordinates to 2D detected objects, and use of intermediate data structures that bridge 2D image space and 3D physical space. These intermediaries enable accurate distance determination while maintaining real-time performance.
2Measurement precision
If complex machine learning models are used for object classification, then the accuracy of object identification improves, but the computational complexity and processing time increase
Solution Approach 1:
The patent segments the object detection process into distinct stages: 2D object detection using simplified models, object classification, 3D reprojection, and distance calculation. This segmentation allows each stage to use appropriately complex methods, with the computationally intensive 3D reprojection only applied to detected objects rather than the entire scene, reducing overall complexity while maintaining accuracy.
Solution Approach 2:
The patent performs preliminary 2D object detection and classification before applying computationally intensive 3D reprojection operations. By pre-identifying objects of interest in 2D space, the system avoids performing complex 3D calculations on every pixel or region, significantly reducing computational complexity while maintaining high classification accuracy for relevant objects.
3Reliability
If multiple sensor types are integrated for distance determination, then the reliability of distance measurement improves, but the device complexity increases
Solution Approach 1:
The patent designs a universal processing framework that can handle multiple sensor types (monocular cameras, stereo cameras, LiDAR, time-of-flight sensors) through a common 3D reprojection and coordinate assignment mechanism. This multi-functional approach allows the system to achieve high reliability through sensor fusion while avoiding the need for separate processing pipelines for each sensor type, thereby controlling device complexity.
Data Source
AI summary
A computer implemented method for determining a distance of an object comprises the following steps carried out by computer hardware components: determining an image containing the object; determining a class of the object based on the image; determining a coarse estimation of the distance based on a distance sensor; and determining the distance of the object based on the coarse estimation and based on the class of the object.