Reference-Object Distance Estimation for Long-Range Targets
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing LiDAR technologies fail to meet long-range sensing requirements for autonomous vehicles, while camera images lack depth information, leading to inaccurate distance estimation of objects beyond their range.
Innovation Solution
A neural network-based system uses reference objects with known distances, obtained from LiDAR or maps, to estimate the distance of long-range target objects by generating pair embeddings and processing them with attention modules to fuse information from different reference objects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If LiDAR sensors are used for distance measurement, then measurement precision is improved, but the sensing range is limited and objects beyond range cannot be detected
Solution Approach 1:
The patent uses reference objects (intermediaries) with known distances to bridge the gap between LiDAR's limited range and camera's lack of depth information. By detecting reference objects within LiDAR range and using them as mediators, the system extends accurate distance estimation to targets beyond LiDAR's direct sensing capability through the neural network's inference from reference object positions and sizes.
Solution Approach 2:
The patent replaces the mechanical/optical direct measurement approach of LiDAR with a computational approach using neural networks. Instead of relying on LiDAR's physical sensing range, the system uses camera images combined with neural network processing to estimate distances of far-range objects by learning from reference objects, substituting physical measurement with intelligent inference.
2Length of stationary object
If camera sensors are used to detect objects, then sensing range is improved, but depth information is lost leading to inaccurate distance estimation
Solution Approach 1:
Reference objects serve as intermediaries that transfer depth information from the LiDAR-measured world to the camera's 2D image space. By detecting reference objects in camera images and matching them with LiDAR distance data, the system creates a bridge that allows the neural network to infer depth information for all objects in the scene, including those beyond LiDAR range.
Solution Approach 2:
The patent transforms the camera's 2D image data into 3D distance information by changing the parameter representation. The neural network learns to map 2D image features (positions, sizes, shapes of reference objects) to 3D distance parameters, effectively converting the camera's strength (wide sensing range) into accurate depth estimation capability through parameter transformation.
3Measurement precision
If neural network processing is applied to fuse reference object data, then distance estimation accuracy for long-range objects is improved, but device complexity increases
Solution Approach 1:
The neural network is designed to perform multiple functions: detecting reference objects, estimating their distances, inferring target object distances, and handling various scene configurations. This multi-functionality reduces the need for separate specialized modules, managing system complexity while achieving accurate long-range distance estimation through a single integrated processing framework.
Data Source
AI summary
Methods, computer systems, and apparatus, including computer programs encoded on computer storage media, for generating a distance estimate for a target object that is depicted in an image of a scene in an environment. The system obtains data specifying (i) a target portion of the image that depicts the target object detected in the image, and (ii) one or more reference portions of the image that each depict a respective reference object detected in the image. The system further obtains, for each of the one or more reference objects, a respective distance measurement for the reference object that is a measurement of a distance from the reference object to a specified location in the environment. The system processes the obtained data to generate a distance estimate for the target object that is an estimate of a distance from the target object to the specified location in the environment.