Trajectory optimisation
Trajectory optimisation improves the path of your survey platform, so point cloud and imagery match each other and the real world. Automated, usually within 24 hours.
A GNSS/INS solution is accurate as long as satellite reception is good. Once reception degrades, the inertial solution takes over and the error grows with the duration of the interruption: a few centimetres at first, and readily decimetres during a longer outage. Multipath along facades and noise barriers also produces positions that report a valid solution while being systematically displaced. Add to that residual errors in the lever arm and the mounting angles, plus the distance to the reference station and the ionospheric conditions at the time of survey. None of this is necessarily visible in the raw point cloud. Only when you overlay two passes, or check a facade against a ground control point, does it become clear that the trajectory has wandered. We therefore start by assessing the quality indicators from the trajectory solution itself, such as the estimated position and orientation uncertainty and the differences between the forward and backward filter passes, and compare those with what the captured data shows.
The captured data is itself the best check on the trajectory. Overlapping strips, for example an outbound and return run or a junction you pass twice, should yield identical geometry. Where they do not, the difference is largely attributable to the trajectory. GEORIZON uses that overlap as an additional observation: planes, edges and other stable features are used to derive correspondences between passes, after which the trajectory is adjusted so that the remaining discrepancies are minimised. Where the route forms a loop, the accumulated drift is distributed across the whole run rather than left as a step at the end. Ground control points or an existing reference dataset are included as an absolute constraint, so that absolute positioning is correct alongside internal consistency. You receive a report with the residuals per control point and the remaining discrepancy between strips, so quality is demonstrable rather than assumed.
Because panoramas, oblique imagery and point cloud are georeferenced from the same trajectory, an improved trajectory benefits every deliverable at once. Measuring in the panorama lines up with the point cloud again, objects from AI detection and vectorisation land in the right place, and successive survey rounds become comparable with one another. Processing is vendor independent: we work from the raw or post-processed trajectory files of common GNSS/INS systems and from the platform's sensor data, whatever the brand. No per-user licence is required and your data stays in your own tenant, with SSO and an on-premises deployment available as options. Once processed, the result is ready in the viewer, including GDPR blurring of faces and number plates, and available through the open API, WMTS and download, delivered in the coordinate reference system your project requires.
