3D Point Cloud Fusion for Resolution-Adaptive Obstacle Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing obstacle detection systems for autonomous vehicles do not effectively account for resolution differences between various 3D sensing devices, leading to inadequate obstacle detection and navigation, especially when one sensor's resolution is poor in certain dimensions or conditions.
Innovation Solution
The method involves generating volumetric surface functions for each 3D point cloud using associated point spread functions to incorporate sensor resolution, forming a composite surface function by multiplying or adding these functions, and applying automated edge-based thresholding to optimize the resolution and detect obstacles accurately.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If sensor fusion is performed by combining object tracks from different sensors, then obstacle detection is achieved, but resolution differences between sensors are not compensated leading to reduced measurement precision
Solution Approach 1:
The patent introduces an intermediary resolution compensation module that processes point clouds from multiple sensors. This module applies resolution-specific transformations to each sensor's point cloud data before fusion, effectively mediating the resolution differences. The compensation uses learned resolution parameters to adjust the spatial distribution of points, allowing high-resolution sensors to enhance the data of lower-resolution sensors in a resolution-adaptive manner.
Solution Approach 2:
The system dynamically changes resolution parameters based on sensor characteristics and operating conditions. By learning resolution parameters from training data and adapting them during operation, the system can optimize the fusion process for different sensor combinations and environmental conditions, thereby improving measurement precision while maintaining reliable obstacle detection.
2Ease of operation
If point clouds are registered and aligned in a common coordinate system, then fusion is achieved, but poor resolution of particular sensors in certain dimensions is not compensated
Solution Approach 1:
The patent applies local quality enhancement by identifying which dimensions or regions of the point cloud suffer from poor resolution and applying targeted compensation only to those areas. Rather than uniformly processing all data, the system adapts the resolution compensation to local needs, maintaining simplicity while improving dimensional resolution where required.
Solution Approach 2:
The system compensates for poor resolution in certain dimensions by leveraging information from other dimensions where the sensor performs better. By transforming and fusing data across multiple dimensions, the system can recover lost resolution information through cross-dimensional correlations, effectively using another dimension to弥补 the deficiency.
3Device complexity
If existing sensor fusion methods are used, then processing is simplified, but resolution differences of sensors are not considered leading to degraded detection performance
Solution Approach 1:
The patent performs preliminary resolution compensation and parameter learning during a training phase before actual operation. By pre-learning resolution characteristics and compensation parameters from training data, the system reduces the computational complexity during real-time operation while maintaining high detection performance. This preliminary action separates the complex learning process from the operational fusion process.
Data Source
Figure 1A
Figure 1B
Figure 2A
AI summary
A method for obstacle detection and navigation of a vehicle using resolution-adaptive fusion includes performing, by a processor, a resolution-adaptive fusion of at least a first three-dimensional (3D) point cloud and a second 3D point cloud to generate a fused, denoised, and resolution-optimized 3D point cloud that represents an environment associated with the vehicle. The first 3D point cloud is generated by a first-type 3D scanning sensor, and the second 3D point cloud is generated by a second-type 3D scanning sensor. The second-type 3D scanning sensor includes a different resolution in each of a plurality of different measurement dimensions relative to the first-type 3D scanning sensor. The method also includes detecting obstacles and navigating the vehicle using the fused, denoised, and resolution-optimized 3D point cloud.