Vehicle LiDAR Dual-Density Point Clouds for Faster Object Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Vehicle lidar systems face increased processing time and complexity with higher point cloud densities, which is undesirable for efficient object detection and vehicle operation.

Innovation Solution

A neural network-based dual density point cloud generator is implemented, reducing point density outside a region of interest (ROI) within the field of view, using a Deep Q-Network (DQN) to optimize the location of the ROI and process the dual density point cloud for object detection, thereby reducing overall processing time and bandwidth requirements.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If point cloud density is increased to improve object detection, then detection accuracy is improved, but processing time and complexity increase

Engineering Contradiction:

Improveobject detection accuracyVSAvoidprocessing time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent applies local quality by creating a dual-density point cloud where different regions have different point densities. The region of interest (ROI) maintains high point density to ensure accurate object detection, while non-ROI areas use reduced point density to minimize processing time. This spatially varying density approach resolves the contradiction by optimizing both detection accuracy in critical areas and processing efficiency in less critical areas simultaneously.

Inventive Principle:

Principle #3Local quality

2Measurement precision

If point cloud density is increased to improve object detection, then detection accuracy is improved, but processing complexity increases

Engineering Contradiction:

Improveobject detection accuracyVSAvoidprocessing complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The system implements local quality by segmenting the field of view into ROI and non-ROI regions with different processing requirements. High-density processing is applied only to the ROI where detection accuracy is critical, while low-density processing handles the rest of the scene. This reduces overall processing complexity while maintaining detection accuracy in the most important regions.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The patent segments the point cloud processing into two distinct density levels based on spatial location. By dividing the field of view into different processing zones (ROI and non-ROI), the system reduces the overall computational complexity compared to uniform high-density processing, while still achieving accurate detection where needed through the high-density ROI processing.

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS11881028B2Vehicle lidar system with neural network-based dual density point cloud generator

Publication Date: 2024.01.23 GM GLOBAL TECHNOLOGY OPERATIONS LLC

AI summary

A vehicle system includes a lidar system that obtains an initial point cloud and obtains a dual density point cloud by implementing a first neural network and based on the initial point cloud. The dual density point cloud results from reducing point density of the initial point cloud outside a region of interest (ROI). Processing the dual density point cloud results in a detection result that indicates any objects in a field of view (FOV) of the lidar system. A controller obtains the detection result from the lidar system and controls an operation of the vehicle based on the detection result.