3D Point Cloud Drivable Area Detection Using CNN Ground Height Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional vehicle drivable area detection methods using depth cameras and three-dimensional point clouds face limitations in accuracy and efficiency, particularly in filtering ground point clouds and determining drivable areas, due to reliance on manual filtering rules and poor generalization abilities.
Innovation Solution
A drivable area detection method utilizing a convolutional neural network-based ground height estimation model to grid three-dimensional point clouds, extract convolution features, perform feature fusion, and determine non-ground point clouds, enabling accurate obstacle detection and drivable area determination.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If manual filtering rules are used to filter ground point clouds, then the filtering process can be simple to implement, but the processing time increases significantly and the generalization ability is poor
Solution Approach 1:
The patent replaces the manual mechanical filtering process with an automated deep learning-based ground height estimation model. The model automatically processes three-dimensional point clouds to estimate ground heights and identify non-ground point clouds, substituting the manual rule-based filtering mechanism with an intelligent system that learns optimal filtering strategies from data, thereby dramatically improving processing speed while maintaining accuracy across diverse environments.
2Ease of manufacture
If manual filtering rules are used to filter ground point clouds, then the filtering logic can be straightforward, but the accuracy decreases for point clouds not covered by the filtering rules
Solution Approach 1:
The patent transforms the filtering approach from using fixed manual rules to using learned parameters from a trained deep learning model. The ground height estimation model learns optimal parameters for identifying ground versus non-ground point clouds from training data, allowing the system to adapt to various terrains and environments. This parameter learning mechanism enables accurate filtering across diverse scenarios without requiring manual rule adjustments.
3Device complexity
If depth camera is used for drivable area detection, then the system structure can be simple, but the detection accuracy is limited by depth perception ability
Solution Approach 1:
The patent transitions from using two-dimensional depth camera data to processing three-dimensional point cloud data for drivable area detection. By utilizing the full three-dimensional spatial information from point clouds and applying ground height estimation in 3D space, the system overcomes the depth perception limitations of depth cameras while maintaining a relatively simple system architecture that builds upon existing point cloud processing capabilities.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The disclosure relates to the field of vehicle technologies, and specifically provides a drivable area detection method, a computer device, a storage medium, and a vehicle, to solve a problem of accurately determining a drivable area. For this purpose, the method of the disclosure includes: obtaining three-dimensional point clouds of a driving environment of a vehicle; estimating a ground height of the current environment based on the three-dimensional point clouds of the current environment by using a ground height estimation model based on a convolutional neural network; determining, based on the ground height, non-ground point clouds not belonging to the ground in the three-dimensional point clouds; performing obstacle detection on the non-ground point clouds to obtain one or more obstacles; and determining a drivable area in the driving environment of the vehicle based on a position of an obstacle. According to the above method, the non-ground point clouds can be quickly and accurately selected from a large number of three-dimensional point clouds, thereby improving the efficiency and accuracy of determining the drivable area by using the non-ground point clouds.