3D Point Cloud Feature Extraction for Static-Dynamic Object Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current digital maps for location-based services lack granularity for precise navigation and route guidance, particularly in distinguishing between static and dynamic objects, leading to inconsistent data and limited usefulness in dynamic environments.
Innovation Solution
A method using point cloud data from LIDAR sensors to compute voxel sequences, extract semantic features, model temporal changes, and classify objects as static or dynamic, employing voxel cloud connectivity segmentation, encoder-decoder networks, three-dimensional convolutional long short-term memory networks, and spatial transformer networks to generate accurate three-dimensional surface models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If digital representations of maps are used for location-based services, then route planning and navigation guidance can be provided, but the maps lack granularity and detail beyond roadways, limiting precision for precise navigation
Solution Approach 1:
The patent segments the environment into discrete three-dimensional objects extracted from point cloud data. Each object is individually identified, classified, and positioned in 3D space, enabling precise navigation by breaking down the continuous environment into manageable, distinguishable units that can be processed and utilized for high-precision location-based services
Solution Approach 2:
The patent transitions from traditional two-dimensional map representations to three-dimensional object models. By adding the vertical dimension and creating volumetric representations of environmental objects, the system achieves superior granularity and detail, enabling precise navigation and route guidance with rich environmental context that 2D maps cannot provide
2Loss of information
If crowd-sourced data and infrastructure monitoring data are used to enhance map detail, then more environmental information becomes available, but the data becomes inconsistent due to dynamic objects that do not persist over time
Solution Approach 1:
The patent implements dynamic object classification by analyzing temporal changes in point cloud data. Objects are classified as static or dynamic based on their movement patterns across multiple time points. This dynamic approach allows the system to adapt to changing environments, maintaining reliable and consistent maps by distinguishing between persistent environmental features and transient moving objects
Solution Approach 2:
The system employs feedback mechanisms by continuously comparing point cloud data across different time points. By analyzing temporal changes and object persistence, the system provides feedback on object stability, enabling it to filter out inconsistent dynamic objects and maintain reliable, consistent environmental representations for location-based services
3Adaptability or versatility
If comprehensive three-dimensional environmental models are created to distinguish static and dynamic objects, then location-based services are enhanced, but the processing complexity and computational requirements increase
Solution Approach 1:
The patent divides the complex task of environmental modeling into sequential processing stages: point cloud acquisition, voxelization, object extraction, classification, and model generation. This segmentation of the processing pipeline reduces computational complexity at each stage while maintaining the ability to create comprehensive three-dimensional models that enhance location-based services with detailed environmental awareness
Solution Approach 2:
The system performs preliminary processing by converting point cloud data into voxel representations before object extraction and classification. This preliminary voxelization organizes the data into a structured format that simplifies subsequent processing steps, reducing overall computational complexity while enabling versatile service capabilities through comprehensive 3D environmental modeling
Data Source
AI summary
Provided herein is a method, apparatus, and computer program product for classifying objects as static objects or dynamic objects based on point cloud data. Methods may include: receiving point cloud data representative of an environment; computing voxel sequences from the point cloud data; extracting voxel-wise semantic features from the voxel sequences; modeling voxel-wise temporal changes based on the voxel-wise semantic features; and classifying objects in the environment as dynamic objects or static objects based on the modeled voxel-wise temporal changes. Computing voxel sequences from the point cloud data may include using a voxel cloud connectivity segmentation method to group voxels in point clouds into perceptually meaningful regions.