Multi-Resolution Pseudo-LiDAR Fusion for Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
LiDAR sensors are expensive and produce sparse point cloud data, limiting their effectiveness in object detection, while image sensors provide denser but less accurate pseudo-LiDAR data with resolution issues affecting object detection accuracy.
Innovation Solution
Generating multiple bird's eye view maps with different resolutions from pseudo-LiDAR point cloud data and combining their features to enhance object detection, addressing the limitations of sparse LiDAR data and resolution-dependent accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If LiDAR sensors are used to detect objects, then measurement precision is improved, but device complexity and cost increase
Solution Approach 1:
The patent creates a pseudo-LiDAR point cloud by copying and transforming image data through depth estimation algorithms. Instead of using actual LiDAR sensors, the system generates synthetic 3D point cloud data from 2D images, achieving LiDAR-like functionality at lower cost and complexity while maintaining object detection accuracy
Solution Approach 2:
The patent replaces the mechanical LiDAR sensing system with a computational approach using image sensors and depth estimation algorithms. The physical laser scanning mechanism is substituted with software-based pseudo-LiDAR generation, reducing hardware complexity while preserving measurement precision for object detection
2Measurement precision
If LiDAR sensors are used, then depth detection capability is improved, but point cloud density decreases
Solution Approach 1:
The patent changes the parameters of image processing by applying multiple depth estimation algorithms with different characteristics (monocular, stereo, multi-view) and combining their outputs. This parameter variation in the processing approach generates denser point clouds while maintaining accurate depth detection capability
3Productivity
If single resolution bird's eye view map is generated, then processing speed is improved, but object detection accuracy deteriorates
Solution Approach 1:
The patent segments the feature extraction process by creating multiple bird's eye view maps at different resolutions (coarse and fine). Each resolution level captures different spatial features, and the segmentation of feature spaces allows the system to maintain high detection accuracy while managing processing load through hierarchical analysis
4Ease of manufacture
If image sensors are used instead of LiDAR, then cost is reduced, but measurement precision deteriorates
Solution Approach 1:
The patent combines multiple depth estimation algorithms (monocular, stereo, multi-view) to create a composite pseudo-LiDAR system. This composite approach integrates the strengths of different algorithms to achieve accurate depth measurement from image sensors, matching LiDAR precision at lower cost
Data Source
AI summary
The embodiments disclosed herein describe vehicles, systems and methods for multi-resolution fusion of pseudo-LiDAR features. In one aspect, a method for multi-resolution fusion of pseudo-LiDAR features includes receiving image data from one or more image sensors, generating a point cloud from the image data, generating, from the point cloud, a first bird's eye view map having a first resolution, generating, from the point cloud, a second bird's eye view map having a second resolution, and generating a combined bird's eye view map by combining features of the first bird's eye view map with features from the second bird's eye view map.