INTERGEO represents today the Global Hub Fair of the Geospatial Community, the leading international trade fair for geodesy, geoinformation and land management.

SITECO will present the new release of  Road-SIT Survey the mobile mapping data application, fully compatible with the most widespread mobile mapping systems like the Optech-Lynx, Riegl-VMX, Topcon IPS3, Leica Pegasus, and of course  Road-Scanner.

Road-SIT Survey, v. 7.0 has been developed in collaboration with the Bologna University  Computer Vision Lab (CVLAB). Smart features have been integrated for the point-cloud classification:

• The Geometric classification which brings big advantages to edge detection and point-cloud editing;

• The Semantic classification that allows a selective view of the point clouds and objects detection through a sophisticated 2D and 3D deep-learning algorithm.

The geometric classification consists in computing Normals and Curvatures in the neighborhood of each point:
• The Normal is the direction of the tangent plane 
• The Curvature is the degree of non-coplanarity: 
Both parameters are of a great help for edge detection and feature extraction.

The user interface has been improved in order to exploit the new features in the best way. In all the views and projections, the point clouds can be displayed in different color ramps based on Intensity, Elevation, True color, Normals, Curvature, Semantic Classification.

For the NORMAL the RGB color ramp is: Blue = vertical (direction Z), Red = horizontal (direction X), Green = horizontal (direction Y)
For the CURVATURE: (GREEN=LOW=plane surfaces, RED=HIGH=edges , corners, natural elements like leaves and terrain)

The new display options are of great help to the feature extraction activities thanks to an real-time representation of the horizontal and vertical surfaces, allowing an immediate identification of the objects to be mapped (roads, furniture, buildings, signage, etc.) from others that can be considered as noise (vegetation, cars, pedestrians).
The curvature value allows the immediate location of the edges and makes it easier to extract them.

The Semantic Classification allows the automatic assignment, to every pixel of the camera pictures, of a LABEL with a class object, out of the 16 recognized classes of the neural network. It is applied to the point-cloud, divided into 5-cm boxes (voxels).

The new features include also the object detection for Signs, Posts, Poles, and Buildings:
• The application detects clusters of adjacent voxels as single objects (poles, signs, buildings);
• Each of them is stored as a new object in a database with its position, picture, and dimensional features; 
• The detected objects can be edited in the feature extraction environment.