3D LiDAR Reflection Point Filtering with Joint Image Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
3D LiDAR scans are corrupted by multiple reflections, leading to incorrect point cloud generation and difficult alignment of multiple scans, with current methods relying on manual removal of reflection points due to the lack of automated semantic perception capabilities.
Innovation Solution
A joint analysis of LiDAR and image data using a comparison algorithm, such as a convolutional neural network, to classify and remove reflection points based on intensity and pattern differences, enabling automated detection and removal of multiple reflections.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If manual removal of reflection points is used, then point cloud quality can be improved, but time consumption and labor intensity increase significantly
Solution Approach 1:
The patent replaces the manual mechanical process of identifying and removing reflection points with an automated computer-based system. The system uses processing units to execute algorithms that automatically detect reflection points by analyzing intensity differences between LiDAR data and image data, eliminating the need for manual human intervention in the point cloud processing workflow.
Solution Approach 2:
The system enables the point cloud processing to be self-sufficient by automatically detecting and removing reflection points without requiring manual human operation. The processing unit autonomously performs the entire workflow of comparing LiDAR and image data, identifying reflection points, and generating cleaned point clouds, making the system self-service capable.
2Productivity
If automated detection methods are implemented, then productivity increases, but system complexity increases due to integration of multiple sensors and processing algorithms
Solution Approach 1:
The system integrates multiple functions into a unified platform: the LiDAR sensor captures distance measurement data, the image sensor captures visual data, and the processing unit performs both data fusion and reflection point detection. This multi-functional integration allows the system to simultaneously achieve automated detection, data correlation, and point cloud generation, improving productivity despite the increased complexity.
Solution Approach 2:
The patent introduces image data as an intermediary element to facilitate the automated detection of reflection points. By comparing LiDAR intensity data with corresponding image intensity data, the system creates a reference framework that simplifies the identification of reflection points. The image data acts as a mediator that bridges the LiDAR measurements and the reflection detection logic, making the automated process more reliable.
3Measurement precision
If multiple sensors are integrated for joint analysis, then detection accuracy improves, but device complexity and cost increase
Solution Approach 1:
The patent merges LiDAR data and image data into a unified analysis framework. The processing unit combines intensity information from both sensors, correlating them through spatial and angular relationships. This merging of data sources enables the system to leverage the complementary strengths of LiDAR (precise distance measurement) and image sensors (visual context), improving detection accuracy while managing integration complexity through systematic data fusion.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Improves point cloud quality by reducing outliers and automating the removal of reflection points, thereby enhancing data acquisition to visualization time and alignment accuracy.
Implementation Method 1
distances are determined based on the so-called pulse transit time method, wherein an emission time and a reception time of an emitted and returning pulse are determined
Implementation Method 2
in case the measurement beam hits a window and is reflected towards the ceiling and then back to the laser scanner
Data Source
AI summary
3D LiDAR scanning for generating a 3D point cloud of an environment, e.g. for surveying a construction side or for building surveying. Data points within the 3D laser scan data originating from multiple reflections within the environment, referred to as reflection points, are automatically identified and removed from the point cloud, by making use of a joined analysis of paired 3D laser scan data and image data captured with a measuring device. Points belonging to objects which appear in the 3D scan data but do not appear in the image data are classified as reflection points. Detection of the reflection points is provided by a reflection point classifier, which is trained to find semantic similarity and dissimilarity of 3D laser scan data and image data.