SLAM 3D Mapping with Real-Time Sensor Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional offshore surveying and mapping technologies, such as MBES, face challenges in achieving accurate and efficient data collection due to positional errors and the need for offline optimization, which can be time-consuming and resource-intensive, especially in harsh environments like the subsea.
Innovation Solution
A computing device equipped with a data processor, range sensor, and SLAM module that synchronizes visual and acoustic data in real-time to generate a combined 3D map, allowing for efficient use of processing power and online optimization, even in resource-constrained environments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If offline optimization is used to correct positional errors and fuse sensor data, then mapping accuracy is improved, but mapping time and computational resources are significantly increased
Solution Approach 1:
The system performs preliminary positioning and mapping actions during the survey itself through SLAM (Simultaneous Localization and Mapping), rather than waiting for offline processing. The computing device continuously estimates its pose and builds the map in real-time as data is collected, enabling accuracy improvements without the time penalty of post-survey optimization
Solution Approach 2:
The patent introduces an intermediary computing device that acts as a bridge between data collection and final mapping. This device performs real-time sensor fusion and pose estimation, mediating the processing burden and enabling accurate mapping without requiring extensive offline computational resources or time
2Productivity
If multiple sensors are integrated for real-time mapping, then mapping speed and coverage are improved, but processing power requirements increase
Solution Approach 1:
The system segments the computational workload by dividing it between the computing device (real-time pose estimation and map building) and offline processing (final map optimization and analysis). This segmentation allows real-time mapping with multiple sensors while managing processing power requirements through distributed computation
Solution Approach 2:
The system performs partial processing in real-time (sufficient for continuous mapping) and reserves excessive or complete processing for offline operations. This approach enables productive real-time mapping without requiring the full computational power to be available continuously, balancing speed and energy consumption
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
This approach enables faster, more accurate, and robust mapping with improved efficiency and reduced costs by integrating visual and acoustic data sources, enabling real-time data gathering and decision-making during surveys.
Implementation Method 1
Conventional offshore surveying and mapping (e.g., bathymetry) is carried out using range sensors such as LIDARs or Multi Beam Echo Sounders (MBES)
Implementation Method 2
receive images from the camera; process the images to develop a SLAM 3D model of the environment
Data Source
Figure 1
Figure 2a~2d
Figure 3
AI summary
A computing device (10) for generating a 3D model of an environment, the computing device comprising: a data processor; a range sensor; a camera; and a simultaneous location and mapping, SLAM, module configured to execute a SLAM 3D reconstruction algorithm to: receive images from the camera; process the images to develop a SLAM 3D model of the environment; estimate pose data for the camera; and identify key frames of the images from the camera, wherein the range sensor has a known spatial relationship with respect of the camera, wherein the data processor is configured to: receive range data from the range sensor; receive pose data from the SLAM module that is associated in time with the range data; determine a first set of range data comprising a first sequence of range data received between a first key frame and a second key frame; and accumulate the first sequence of range data into a first range sub-model based on a first sequence of pose data received between the first key frame and the second key frame.