MLF-SLAM
No single sensor has to carry the whole trajectory.
MLF-SLAM fuses LiDAR, visual cameras, and IMU into one localization solution, so a PRECISE scanner keeps tracking accurately even when one sensor type runs out of usable signal: open fields with few visual features, long uniform corridors with few geometric features, or underground spaces with no GNSS at all.
The Fusion Principle
Why a Single Sensor Isn't Enough
Every SLAM (Simultaneous Localization and Mapping) system needs a continuous stream of distinguishing signal to know where it is from one moment to the next. A single-sensor SLAM system depends entirely on one type of signal: visual SLAM needs distinct visual features, LiDAR-only SLAM needs distinct geometric structure. When that signal runs out, tracking drifts:
Open fields, beaches, and reclamation areas
Few distinct visual or structural features for a camera or a LiDAR-only system to lock onto.
Long, uniform corridors
Repeat the same geometric shape for meters at a time, which can confuse geometry-only tracking.
Underground parking structures, tunnels, and mines
No GNSS signal at all, ruling out satellite-based position correction entirely.
MLF-SLAM's answer is architectural, not statistical: instead of trying to make one sensor more robust, it fuses multiple independent sensor streams (LiDAR, visual cameras, and IMU) into a single localization solution, so when one sensor's signal runs out, the others still have something to anchor the trajectory to. None of these sensors depend on satellite signal, which is what lets the whole system operate independently of GNSS.
In PRECISE Products
MLF-SLAM Across the Scanner Line
The Full MLF-SLAM Implementation
PRECISE S7
S7's proprietary MLF-SLAM engine fuses three sensor streams into one localization solution:
- +LiDAR (HMid360 sensor, 200,000 points/second, Class 1 eye-safe, 905 nm, 360° horizontal FOV / −7° to 52° vertical FOV) for geometric structure.
- +Dual 1.3 MP visual SLAM cameras for visual tracking reference alongside the LiDAR geometry.
- +IMU (200 Hz output) for motion and orientation data between sensor updates.
This is what lets S7 maintain stable positioning in the weak-feature and GNSS-denied environments above. A separate onboard RTK module then anchors the SLAM-generated relative geometry to absolute real-world coordinates when georeferenced output is needed: MLF-SLAM handles local tracking, RTK handles the coordinate anchor.
5 cm absolute (real-time) · under 3 cm absolute (post-processing) · under 2 cm relative
Explore PRECISE S7The Same Fusion Principle, Tuned for Field Speed
PRECISE S2
S2 takes a related but distinct approach, built around single-operator speed rather than S7's full capture-to-deliverable pipeline. It combines:
- +360° hybrid rotating-mirror LiDAR, which progressively builds a full-surround field of view (PRECISE reports S2 reaches near-total spatial coverage within about 1 second of continuous operation), mounted at a fixed 25° downward tilt to capture floor, forward objects, and ceiling in a single pass.
- +A visual SLAM (V-SLAM) camera contributing an additional tracking reference alongside the LiDAR.
- +A 6-axis, 200 Hz IMU (3-axis accelerometer + 3-axis gyroscope) for high-frequency motion tracking that supports stable scan quality during active handheld movement.
S2 also includes its own onboard RTK module (0.8 cm + 1 ppm horizontal / 1.5 cm + 1 ppm vertical) running on the MATRIX positioning engine, for teams that need the resulting point cloud tied to absolute coordinates.
Point cloud thickness ≤1 cm · relative accuracy ≤2 cm · absolute accuracy ≤5 cm
Explore PRECISE S2Choosing between them: S7 and S2 apply the same multi-sensor fusion logic to different priorities. S7 is built for a complete capture-to-professional-deliverable workflow with a deeper processing pipeline; S2 is built for fast, single-operator field capture with real-time true-color point cloud confirmation. Both avoid single-point-of-failure tracking; the difference is in the surrounding workflow, not the fusion principle itself.
FAQ
Common Questions
Multiple Localization Fusion-SLAM: PRECISE's proprietary framework for fusing multiple sensor types (LiDAR, visual cameras, IMU) into a single localization solution, rather than depending on any one sensor type to maintain tracking.
No. MLF-SLAM operates independently of GNSS, which is why PRECISE scanners built on it can operate in underground parking structures, tunnels, and other GNSS-denied spaces. GNSS/RTK is used separately, when needed, to georeference the resulting point cloud to absolute coordinates; it isn't required for the SLAM tracking itself.
Because each sensor type depends on a different kind of environmental signal: cameras need visual features, LiDAR-only systems need distinct geometry. In environments where one of those is sparse (an open field for visual features, a long uniform corridor for geometry), a single-sensor system loses its anchor and drifts. A fused system still has another sensor stream to hold the trajectory steady, so no single weak environment defeats the whole tracking solution.
They share the same underlying principle (never depend on one sensor alone), but PRECISE uses the "MLF-SLAM" name specifically for S7's engine, built for a full capture-to-deliverable pipeline. S2 runs a related three-sensor fusion tuned for fast single-operator capture; it isn't branded as MLF-SLAM. See the S7 and S2 breakdowns above for the specific sensors each one uses.
Scanning/mapping accuracy is reported separately from any RTK fix-rate figure. S7: 5 cm absolute accuracy in real-time, under 3 cm post-processing, under 2 cm relative. S2: point cloud thickness ≤1 cm, relative accuracy ≤2 cm, absolute accuracy ≤5 cm. These numbers describe the SLAM-generated point cloud itself, separate from the onboard RTK module's GNSS fix-rate statistics.
Environments where single-sensor SLAM or GNSS alone would struggle: open fields, beaches, and reclamation areas (weak visual/geometric features); long corridors and repetitive structures (weak geometric distinction); and underground parking structures, tunnels, mines, and other GNSS-denied spaces.
Have a site where other scanners lose tracking?
Tell us about the environment: open terrain, underground structure, or a mix, and a PRECISE engineer will help you determine whether S7's full processing pipeline or S2's field-speed workflow fits your project.