Monocular SLAM Initialization Using Pre-Acquired Environmental Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Monocular SLAM systems face challenges in initializing camera pose estimation and determining metric scale factors, particularly in hand-held AR applications, where users struggle to perform distinct camera movements and rotation-only motion results in degenerated models with undetermined scale factors.
Innovation Solution
A method that generates a geometrical model of a real environment using image information from a mobile device's camera, where environmental data acquired by sensors of a mobile system, such as a car, is used to create an initial model, allowing for tracking and subsequent model updates without requiring distinct camera movements, and employing depth information from sensors like range sensors or time of flight cameras to determine correct scale factors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If monocular SLAM is used for camera pose estimation, then the system can operate without pre-knowledge of the environment, but the initialization requires distinct camera movements which are difficult for users to perform and rotation-only motion produces degenerated results
Solution Approach 1:
The patent applies preliminary action by pre-acquiring environmental data with a mobile system (car, drone, etc.) before the user arrives at the location. This pre-collected data is stored in a database and later used to initialize the SLAM system, eliminating the need for users to perform complex initialization movements with their handheld devices.
Solution Approach 2:
The patent introduces an intermediary mobile system (such as a car or drone) that collects environmental data independently. This intermediary acts as a mediator between the environment and the user's SLAM system, transferring pre-processed geographical information to the user device without requiring the user to perform data collection themselves.
2Device complexity
If monocular camera is used for SLAM, then the device is simple and portable, but the metric scale factor remains undetermined leading to inaccurate overlay of virtual information
Solution Approach 1:
The mobile system serves as an intermediary that provides metric scale information to the monocular SLAM system. The mobile system's sensors (odometry, GPS, inertial sensors) measure real-world distances and scales, which are then transferred to the handheld device to enable accurate scaling without adding complexity to the user's camera system.
Solution Approach 2:
The mobile system performs preliminary measurement of metric scale factors during its traversal of the environment. These pre-measured scale values are stored and later applied when the user's SLAM system processes images, allowing the simple monocular camera to achieve metric-scale accuracy through pre-acquired scale information.
Data Source
AI summary
A method of tracking a mobile device comprising at least one camera in a real environment comprises the steps of receiving image information associated with at least one image captured by the at least one camera, generating a first geometrical model of at least part of the real environment based on environmental data or mobile system state data acquired in an acquisition process by at least one sensor of a mobile system, which is different from the mobile device, and performing a tracking process based on the image information associated with the at least one image and at least partially according to the first geometrical model, wherein the tracking process determines at least one parameter of a pose of the mobile device relative to the real environment. The invention is also related to a method of generating a geometrical model of at least part of a real environment using image information from at least one camera of a mobile device.