Lidar Point Cloud Segmentation for Reliable Autonomous Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional algorithms for lidar point cloud segmentation in autonomous vehicles are prone to errors due to developer bias and are not robust to variations in the driving environment, making it difficult to accurately identify and distinguish between similar objects.
Innovation Solution
A neural network-based lidar data segmentation system that processes lidar point cloud data to assign labels to points, determining whether they belong to the same object by computing distance values between output vectors, thereby improving object identification accuracy and reducing segmentation errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If conventional rule-based algorithms are used for lidar point cloud segmentation, then the system is easier to implement and understand, but the accuracy of object identification deteriorates due to developer bias and inability to handle variations in driving environment
Solution Approach 1:
The patent replaces conventional rule-based algorithms (mechanical/systematic approach) with a neural network-based machine learning system. The neural network learns segmentation patterns from training data rather than relying on manually crafted rules, thereby improving object identification accuracy while eliminating developer bias. The system processes lidar point cloud data through learned features and distance computations to achieve more robust segmentation.
2Device complexity
If conventional rule-based algorithms are used for lidar point cloud segmentation, then the system complexity is lower, but the reliability of object identification deteriorates due to errors in distinguishing similar objects
Solution Approach 1:
The patent substitutes rule-based algorithms with a neural network-based system that learns from data. The neural network processes lidar points by computing features and distance values between output vectors, automatically learning robust segmentation criteria that improve reliability in distinguishing similar objects in various driving environments.
Solution Approach 2:
The patent changes the approach from fixed rule-based parameters to dynamic learned parameters. The neural network learns optimal segmentation parameters from training data, including feature extraction parameters and distance threshold parameters, allowing the system to adapt to different driving scenarios and improve reliability.
3Measurement precision
If a neural network-based segmentation system is used, then the accuracy of object identification is improved, but the device complexity increases
Solution Approach 1:
The patent segments the lidar point cloud processing into distinct stages: feature extraction, neural network classification, distance computation, and label assignment. This segmentation of the processing pipeline manages complexity by organizing the neural network system into modular components, each handling a specific aspect of the segmentation task.
4Productivity
If conventional algorithms are used for lidar segmentation, then the processing speed is faster due to simpler computations, but the accuracy of distinguishing similar objects deteriorates
Solution Approach 1:
The patent performs preliminary feature extraction and neural network classification before final segmentation decisions. By pre-processing the lidar data through learned features and computing distance values in advance, the system prepares optimized representations that enable accurate object identification while maintaining efficient processing through cached intermediate results.
Data Source
AI summary
An autonomous vehicle is described herein. The autonomous vehicle includes a lidar sensor system. The autonomous vehicle additionally includes a computing system that executes a lidar segmentation system, wherein the lidar segmentation system is configured to identify objects that are in proximity to the autonomous vehicle based upon output of the lidar sensor system. The computing system further includes a deep neural network (DNN), where the lidar segmentation system identifies the objects in proximity to the autonomous vehicle based upon output of the DNN.