Visual Positioning and VIO
Inside a large exhibition hall, a visitor uses a smartphone while vision recognizes the space and connects the camera to a reference map. VIO combines successive images with inertial data to estimate motion, position and orientation along the route.
Recognize the place, follow the motion
The scientific tutorial on Visual Place Recognition defines the task as recognizing previously seen places from images. Recognition depends on what the camera sees, the overlap between viewpoints and changes in the appearance of the environment.
Visual Positioning uses that connection with a prepared environment to determine where the camera is and how it is oriented. Visual-Inertial Odometry answers a temporal question instead: how the device moved relative to the previous instant, combining the visual sequence with accelerometer and gyroscope data. The two estimates can cooperate, but they are not interchangeable.
The reference is prepared first; the live camera is then localized against the model.
From mapping to position and orientation
The InLoc research on indoor visual localization describes estimating the six-degree-of-freedom pose of a photograph relative to a three-dimensional map of the environment. The method compares the current image with scene references and addresses changes in viewpoint, layout and occlusion.
In the GiPStech portfolio, Visual Mapper collects reference images while an operator walks through the site and contributes to building the visual model. During operation, the live camera observes the scene, extracts recognizable features and connects them to the spatial reference to estimate position and orientation. Mapping comes before localization: avoiding beacons does not remove the need to prepare the environment.
Two technologies with distinct, complementary roles
ORB-SLAM3 documents an architecture that can work with visual and visual-inertial sensors, reuse maps and handle relocalization. It is an open-source research system used here to explain general principles, not GiPStech implementation or performance.
Visual Positioning seeks a connection with a spatial reference that has already been built. VIO integrates change observed in the images with IMU measurements. Position and orientation result from geometry and motion estimation, while a recognized reference can bring the trajectory back into a known coordinate system.
Visual recognition
Matches live-camera features against the environment model to identify a place that has already been represented.
Image sequence
Changes between frames describe how the camera moves relative to the observed scene.
Accelerometer
Measures specific acceleration and contributes to estimating dynamics and scale together with visual observations.
Gyroscope
Measures angular velocity and helps track changes in orientation, including during rapid motion.
Resulting pose
Combines position and orientation in an estimate consistent with the reference and recent motion.
Relative continuity, drift and relocalization
Research on visual-inertial relocalization makes the distinction clear: VIO can be locally accurate, but it remains relative to its starting frame and can accumulate drift. Recognizing a scene in a previous map can recover a pose in the known reference and contribute to global correction.
Within the scope described by the portfolio, Visual Positioning and VIO can therefore support each other: VIO maintains continuity between recognitions, while a new reliable connection with the map can constrain accumulated error. Actual behavior depends on the map, calibration, lighting, motion and the quality of observed features. Possible relocalization does not mean instant relocalization in every circumstance.
Prepare the environment and keep it recognizable
The benchmark on visual localization under changing conditions shows that changes in illumination, season, geometry and viewpoint can challenge established methods. Its results concern specific research datasets and systems, but confirm that the visual quality of the environment must be assessed rather than assumed.
Spaces with stable details, recognizable texture and adequate light support visual matching. Uniform or repetitive surfaces, backlighting, darkness, occlusion, abrupt motion and major changes can reduce the information available. Image collection, retention and updates also need rules consistent with the project's purpose, security and privacy requirements.
Assess the site
Check lighting, distinctive features, repetitive areas, occlusions, scene dynamics and how the camera will be used.
Build the reference
Collect images and coordinates through a controlled survey, prepare the model and document map version and coverage.
Validate in the field
Test devices, motion, viewpoints and real conditions; measure acquisition, continuity and relocalization within the agreed scope.
Maintain the map
Define when to update the reference after changes to layout, furnishings, lighting or materials, and how collected images are governed.
Devices, integration and application contexts
The GiPStech Technology page places vision, inertial sensing and other sources within a modular portfolio. This page examines the visual and visual-inertial branch; Sensor Fusion remains the wider architecture that coordinates different sources.
The GiPStech retail solution documents location-aware services delivered through smartphones, including indoor navigation and contextual content. Visual Positioning and VIO require a device with accessible, compatible camera and IMU sensors; availability must be verified for each model, operating system and usage mode.
The airport solution page shows how position, maps and applications can support wayfinding and operations in a complex infrastructure. Localization technology remains distinct from the operational layer and must be validated for the specific airport.
The public-spaces solution connects localization, wayfinding and contextual content in museums, fairs and events. Augmented reality can be an application enabled by pose, but anchoring digital content belongs to the AuRea page and is not developed here.
RF and markers can provide complementary references under different conditions; GNSS extends the framework outdoors. Continue with Adaptive Hybrid Localization, Augmented Reality Localization and AuRea, and Outdoor and Vehicle Localization.
Smartphone and live camera
The camera observes the environment and the IMU measures motion. Device posture, power use, lighting and continuous operation must be tested in the real service.
Device requirements
Visual Positioning and VIO require camera and IMU access for the localization engine. Device model, operating system, calibration and camera mode require technical confirmation.
Location-aware integration
Position and orientation can feed navigation and contextual services. SDKs, APIs, latency and offline operation must be confirmed for the current scope.
Environments to assess
Retail, airports, exhibition halls, museums, campuses and industry can benefit from vision when the scene, device and requirements allow it.
Assess the environment, device and continuity requirements
Describe the space, target, visual conditions, acceptable mapping work and position and orientation requirements. We can define a field-verifiable trial.
