Pseudo-LIDAR Point Clouds with Surface Normals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current vehicle sensors, such as LIDAR, face challenges including high cost and sparseness of data, while camera sensors provide dense but feature-rich data lacking inherent depth information, making object detection and tracking less accurate.
Innovation Solution
A method and system that generate point clouds with surface normal information by converting camera images into depth maps and then into pseudo-LIDAR point clouds, incorporating surface normal data to enhance object detection and tracking accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If LIDAR sensors are used to generate point clouds, then measurement precision and reliability are improved, but cost increases and data sparseness occurs
Solution Approach 1:
The patent creates a pseudo-LIDAR point cloud by copying depth information from camera images through depth map generation. Instead of using actual LIDAR sensors, the system replicates LIDAR-like point cloud data from camera images, achieving similar measurement precision without the cost and data sparseness issues of real LIDAR systems
Solution Approach 2:
The patent replaces the mechanical LIDAR sensing system with a camera-based computational approach. By substituting the optical ranging mechanism of LIDAR with camera image processing and depth estimation algorithms, the system achieves comparable depth measurement precision while providing dense data coverage
2Quantity of substance
If camera sensors are used to capture images, then cost decreases and data density increases, but depth information is lost
Solution Approach 1:
The patent transforms two-dimensional camera image data into three-dimensional depth information by generating depth maps. This dimensional transformation allows the system to extract depth values from image pixels, creating pseudo-three-dimensional point cloud data that preserves the dense data characteristics of cameras while adding depth measurement precision
Solution Approach 2:
The patent changes the parameter representation of image data by extracting depth values from image pixels. By transforming the parameter space from two-dimensional color/intensity values to three-dimensional coordinates with depth values, the system recovers depth information while maintaining the dense data structure of camera captures
3Measurement precision
If surface normal information is added to point clouds, then object detection accuracy is improved, but device complexity increases
Solution Approach 1:
The patent performs preliminary computation of surface normal vectors during the depth map processing stage, before the point cloud is used for object detection. By pre-calculating and storing surface normal information alongside depth values, the system improves detection accuracy without adding complexity to the real-time detection pipeline
Solution Approach 2:
The patent introduces surface normal vectors as an intermediary element that bridges depth map data and object detection tasks. These normal vectors serve as additional geometric features that enhance the point cloud representation, providing orientation information that improves detection accuracy without requiring fundamental changes to the detection architecture
Data Source
AI summary
A system for generating point clouds having surface normal information includes one or more processors and a memory having a depth map generating module, a point cloud generating module, and surface normal generating module. The depth map generating module causes the one or more processors to generate a depth map from one or more images of a scene. The point cloud causes the one or more processors to generate a point cloud from the depth map having a plurality of points corresponding to one or more pixels of the depth map. The surface normal generating module causes the one or more processors to generate surface normal information for at least a portion of the one or more pixels of the depth map and inject the surface normal information into the point cloud such that the plurality of points of the point cloud include three-dimensional location information and surface normal information.