Version française

Version 36.1 arrives like a toolbox upgrade for power users. Precision tightens. Control deepens. Automation becomes negotiable. Let's walk through what's new:

Utility workflows now move from observation to actionable compliance.

• Automatic, high-precision vectorization of poles and powerlines

• Interactive tools to manually adjust and refine extracted lines

• Fully editable geometries

Users can now define their own compliance profiles:

• Pole tilt threshold

• Minimum line-to-ground clearance

• Clash detection tolerances

Think of it as giving engineering standards a digital measuring tape.

• Flags poles exceeding defined tilt angle

• Generates detailed reports including:

  • Geographic location

  • Measured angle

  • Area map visualization

No more visual guessing. Every leaning pole is quantified.

• Computes ground clearance per segment

• Highlights non-compliant sections

• Produces map-based report with flagged segments and precise locations

It transforms a long transmission corridor into a color-coded compliance dashboard.

• Detects conflict zones between vegetation and powerlines

• Identifies areas requiring trimming

• Generates detailed report with attached map

Maintenance planning now starts with data, not field surprises.

The "Training a New Model" feature now runs on the VisionLidar 2.0 engine. In this version we have made available the retraining of the models generated by this engine.

Customers can:

• Retrain models using their own datasets

• Fine-tune pre-trained architectures

• Adapt models to specific project realities

• Add new classes

• Remove unnecessary ones

• Align class structure with imported PLY content

Your model no longer dictates your workflow. Your workflow shapes the model.

A new entry in the Analyze section: Line Detection.

🔍 Linear Object Extraction

• Detects and vectorizes any linear object in the 3D scene

  • Road markings

  • Cables

  • Edges

  • Structural lines

• Applies geometric simplification

• Produces cleaner, straighter lines

• Reduces unnecessary vertices

• Export vectorized lines to desired formats

• Integrate directly into CAD, GIS, or BIM workflows

From dense points to disciplined polylines, in a few clicks.

We are proud to introduce VisionLidar 2.0, a new deep-learning engine built on transformer technologies inspired by ChatGPT-class architectures.
This new generation delivers:

• Up to 10× faster inference

• Up to 10× more memory efficiency

• Up to +10 points increase in classification accuracy

• Three new general-purpose models optimized for:

  • GeoPlus Mobile

  • GeoPlus Aerial

  • GeoPlus Indoor

This breakthrough dramatically accelerates workflows while improving precision across all major LiDAR acquisition types.

The AI Classification experience has been completely redesigned using a modern QT user interface for improved clarity and usability.
New functionalities include:

• Ignore Class option: exclude selected classes during classification to prevent unwanted assignments.

• Enhanced dataset visualization and improved workflow ergonomics.

• Faster model loading and smoother user interaction.

A fully modernized Train a Model interface is now available, featuring:

• DNN Project Management: automatically stores the PLY files used for training, enabling future re-training, comparison, or auditability.

• Embedded PLY Viewer: inspect point clouds before training, adjust class numbers, names, and colors directly in the interface.

• Dedicated Logs & Results Tab: track training progress, metrics, and outputs with much more transparency and control.

This module lays the foundation for advanced customizable AI pipelines inside VisionLidar.

VisionLidar V36.0 now reads and exports additional LAS/LAZ attributes, critical for aerial workflows:

• Point Source ID

• Reflectance

• Number of Returns

• Echo

This ensures richer datasets, better analytics, and improved compatibility across third-party platforms.

Utilities Analysis has been significantly enhanced:

• Parallelized processing for faster results on large datasets

• Improved flexibility to handle aerial/mobile data with varying densities

• Higher reliability on complex utility projects

Road marking vectorization receives a major upgrade:

• Train a VisionLidar 2.0 model to classify road markings in your project.

• Use the new Detect by Class functionality in the Road Marking dialog to auto-vectorize markings directly from AI outputs.

This dramatically improves automation for transportation, civil engineering, and municipal scanning workflows.

A large portion of V36.0 includes extensive backend improvements to increase:

• Stability

• Scalability

• Performance

• Code maintainability

These upgrades ensure a robust foundation for the 2026 development roadmap.

New Mesh Generation

• A completely redesigned mesh generation feature, featuring both a new methodology and a modernized user interface. This update provides greater control, enhanced output quality, and a more intuitive workflow.

Redesigned "Road by Section" Tool

• The Road by Section module has been fully revamped with a new UI and workflow, making road modeling more efficient and user-friendly.

Edge Detection Overhaul

• Edge detection now benefits from a fresh UI and structural redesign, resulting in clearer results and easier workflow.

New Road Marking Detection Methodology (First Iteration)

• This version introduces the first iteration of a new methodology for road marking detection, designed to deliver improved detection accuracy and adaptability to various environments.

Build System Modernization

• The internal build system has been migrated to CMake and Qt6, streamlining development, improving maintainability, and simplifying future portability.

• Faster processing speeds for mesh and DTM generation;

• Improved stability, preventing crashes when handling large point clouds.

LGSx Format

• Full support for the LGSx format for both point clouds and spherical images, ensuring seamless integration and enhanced compatibility with the new Leica data format.

Improvements in Automatic Utilities Analysis and Vectorization

• Refinements to the automatic Utilities Analysis and vectorization features provide more accurate results and improved processing efficiency, making power line and pole detection faster and more reliable.

Bug Fixes in AI Models and DNN Feature

• Several bugs have been fixed in the AI models and deep neural network (DNN) feature, improving stability and ensuring more consistent outcomes for training and classification tasks.

Redesigned Road Mark Detection

• The road mark detection feature has been given a fresh design, resulting in better usability.

Improved performance and stability

• Internal code improvements and optimizations have been made to maximize overall performance and reliability.

This release delivers improved user experience, enhanced workflows, and faster processing for LiDAR data classification.

‌

Enhanced "New Project" Dialog with a New Design

• A complete redesign of the "New Project" dialog provides a more intuitive and organized user interface for project creation. The improved layout includes better tab organization for settings and parameters. A separate tab is dedicated to uploading images. Ability to add survey dates for the project along with bug fixes related to project creation.

Improved "Spherical Images" Dialog with a new design

• The "Spherical Images" dialog has been revamped for a more streamlined user experience. It now allows users to add spherical images to LiDAR projects without the need to store them in a specific folder. Additionally, users can now view spherical images independently, without requiring associated point clouds. Multiple bugs related to spherical image addition have also been fixed.

Optimized Utilities Analysis & vectorization with a new design: Significant improvements have been made to the Utilities Analysis and vectorization process:

• User-Friendly, Simplified UI: The updated interface is more intuitive and easier to navigate.

• Scene-Based Point Editing: Users can now edit points directly in the scene (move or delete) rather than relying on the table, offering greater flexibility and control.

• Enhanced Line Detection and Vectorization: The line detection algorithm has been enhanced for better accuracy in identifying lines, improving the overall classification process.

Faster Classification for Large Projects

• A new data partitioning method accelerates the data preparation process for the DNN module, significantly reducing classification times for large projects.

Real-Time Training Results Visualization

• Users can now view training and validation graphs in real-time as the model trains, providing greater insight into model performance during training.

Dataset-Specific Profile Selection for AI Model

• New options to select the appropriate profile (Indoor, Mobile, Aerial, or UAV) for your dataset, ensuring that DNN module settings are optimized for the specific data type.

AI Model Training Time Optimization

• The training process is now more efficient, with the ability to stop training early using the validation mIoU to detect when the model is no longer improving over multiple epochs.

New High-Performance Indoor and Outdoor AI Models

• New AI models for indoor and Outdoor classification are now available, offering high performance out of the box. These models are ready for immediate use or can be retrained with custom data.

‌

General

• Resolve case sensitivity problem when importing spherical images

• Optimize point cloud selection to improve performance and reduce resource usage

‌

General

• Export to Autodesk ReCap format (.rcp), including spherical images.

• Ability to create a Surface from selection or fence.

• Spherical images: Units and projection conversion.

• Multiple user experience improvements.

Analysis

• Catenary/Power Lines: Enhanced detection and vectorial object generation.

• Catenary/Power Lines: Span view and distance to ground analysis.

• Catenary/Power Lines: Interactive edition and productivity increase.

• Road Marking: Enhanced detection and vectorial object generation.

Deep Learning

• New PyTorch implementation with general performance boost.

• Ability to resume, fine-tune, and create models based on pre-trained models.

Tools

• Images-to-scan: Map view, image selection and streamlined user experience.

Several corrections have also been applied to this version.

‌

General

• Export Object panel elements (surfaces, meshes, cylinders and trees) in DWG format from Autodesk.

• New CAD viewer with "Realistic view" and "3D Orbit" navigation option when exporting DWG file format.

• Enhancement on camera position and view direction on VisionLidar map interface.

• New class grouping module allowing to organise your classes and create named groups with class number of your choice providing a better use of the full 256 class range.

• New scan grouping module allowing to organise scans with named groups.

• Enhancement on the Fence tool to better keep fences position in a project when changing computers.

• Enhancement on the IFC (BIM) module to have access to VisionLidar internal elements.

• Compatible with new SDK Faro LS 1 1.905.1 (for FLS, LSPROJ and Workspace) when creating a new project.

• Export VisionLidar class definition when saving in RCP format using new Autodesk Recap class definition.

View

• Compatible with the spherical images "Applanix.txt" positioning document from the Applanix mobile platform.

• Updated Viametris spherical images file position allowing the usage of their new multiscan equirectangular jpeg names output.

• Enhancement on spherical images view when combining point cloud and bubble view as well as Image disks.

Analysis

• Improved "Analyse Section" function, allowing finding, and creating sections by elements on a class along an alignment path.

• Improved "Analyse Profile" interface.

Several corrections have also been applied to this version.

General

• New external object insertion module for greater control (point of origin, units, etc.) over inserted 3D elements such as ".OBJ" or ".PLY".

• New tool for creating the colored mesh from the point cloud color (under Surface tab).

• Export colored mesh in "PLY" format.

• Possibility of transferring a mesh from the "Surface" tab to the "Inserted objects" tab, allowing you to modify the object's origin and move it within the scene.

• RCP format export option allowing the usage of scan concept to distribute point cloud from each class and thus have control of classes on Autodesk products.

• Addition to the list of sensors, several camera models compatible with the "Image to scan module".

View

• Extract spherical images from E57 exported from the IVION platform from NavVis mobile solution.

• Compatible with the spherical images "pano-poses-registered.csv" positioning document from the NAVVIS mobile platform.

Analysis

• Improved "Analyze Section" function, allowing you to create sections between two points, according to the nodes of an alignment, or by analyzing point cluster on a class.

Deep Learning

• Use of partitions when directly classifying LAS files using AI (from New Project or Classify LAS).

• Use of corridors when classifying by DL which allows to process only desired areas.

Several corrections have also been applied to this version.

‌

General

• New option allowing us to apply our powerful Neural Network classification module while creating a new project coming from LAS scans format ("New project" function).

• New option to create colored meshes from the point cloud.

• Option to copy or move the mesh or surface to "Inserted objects" allowing the relocation of the mesh over the 3D space .

• Set up the camera position view to improve the visualization of shading on surfaces.

• Rapid inversion of the surface normals.

• Import surfaces from DXF file format.

• Improved projection and unit inputs during project creation.

• Creation of a contour line delimiting the project on the map and on VisionLidar365

View

• New display function by normals which can be applied when point clouds have neither intensity nor RGB display information.

• Improved display for mesh shading areas.

Analysis

• Option to transfer Cogo points or lines with Feature Codes from the following functions: "Cluster detection", "Utilities Analysis", "Edge detection", "GuardRails detection" and "Pavement Markings".

• Option to transfer to Cogo type points with Feature Code, Trees or Cylinders type objects.

Deep Learning

• Improved interface for visualizing results from processed or unprocessed partitions.

• Enhanced version of the events reported in the Log Section.

• Option to classify within a selected corridor.

• New Deep Learning classification batch process made directly on LAS files allowing to keep third party attributes intact.

Several corrections have also been applied to this version.

‌

General

• Registration by list now possible with the point cloud format compatible with VisionLidar 365.

• Point-to-point registration now possible with point cloud format compatible with VisionLidar 365.

• Point-to-point registration with external file now possible with point cloud format compatible with VisionLidar 365.

• The "Move Origin" option is now enabled with point cloud format compatible with VisionLidar 365.

• New option to add several OBJ-type volumetric object files to the 3D scene via a position file.

Edit

• Eraser function is now possible with the point cloud format compatible with VisionLidar 365.

• Classifier function is now possible with the point cloud format compatible with VisionLidar 365.

View

• Improved the function of coloring the point cloud project by spherical images with the addition of new parameters.

Analysis

• Improvement of the Aerial Classification tool in order to clean up some residues of the classification by Deep Learning.

Survey

• Addition of a function to update null elevations (equal to zero) on the Cogo points attached to your database. The function will take the average of the elevation of the active point cloud (class(es) displayed) at the location of the Cogo points (x y) and insert this value in the z field when it is equal to 0. Any Cogo points that already have an elevation value will not be impacted.

• New 3D Land Parcels visualization function allowing you to select and graphically represent on your point cloud (and the location map), any Parcels present in your database.

Deep Learning

• New interface adding a graphical representation of partitions.

• Chromatic representation of partitions already processed.

• Addition of a "Log" section describing processing status.

Settings

• New tool allowing to update certain fields of the original LAS file(s) being used to create a VisionLidar project. This function allows you to select the attributes to update such as, 3D coordinates, RGB color or classes and connect to the original file and replace them with the new values processed by the different Vision Lidar functions such as classification by Learning deep.

‌

General

• New Format (2022) compatible with the VisionLidar 365 a Geo-plus secured Web platform for point cloud solutions. VisionLidar remains compatible with previous local format (.vps). No duplication data for the Web or desktop platform anymore.

• Simplified project creation interface allowing "Drag and Drop" of raw files.
  New interface to visualize EPSG or ESRI projection code entered.

• 256 classes are now supported (same as LAS 1.4).

• New option to regroup files when merging scans during project creation.

• Export LAS 1.4 file format with 256 classes.

Edit

• Enhancement on the Fence tool Starting to pick when a new fence is created. New option to select inside the fence or export the fence area data for training.

Analysis

• Edge detection is now part of VisionLidar Premium

• Pavement Markings detection is now part of VisionLidar Premium

• New option to define specific boundaries on Planimetric Orthophoto function.

DNN

• Progress bar implemented for Deep Learning classification, for better tracking of the progression for the overall process.

• Multi progress bar available for computers with multi-GPU cards allowing monitoring process progression for each individual Card.

• Progress bar implemented for Deep learning training module.

• Log created to monitor system processing.

Settings

• Projection is now described in addition to the Code.

‌

General
• Correction when importing intensity for PTS and CSV type files.
• Optimized selection of the projection code now listed by EPSG code or ESRI code.
• Export of visible classes to the VisionLidar365 format.
• Correction when points are deleted, they are no longer exported.
• Export to Potree 2.0 format, thus increasing the speed of exporting and concatenating files into a few binary documents.
• Ability to export VisionLidar365-viewer.exe viewer when exporting to VisionLidar365 format.
• Significant increase in performance when creating projects by using the Multiprocessor option (Multicore).
• Possibility of simplifying the density of the point cloud from the creation of the project.
• New import format (Workspace "* .fws") for file from FARO systems. This format is in addition to the already existing format "* .lsproj" and "* .fls.
• Ability to export pole vectorization when no power line is vectorized.

Tools (new menu)
• 3D representation of the results from the Volume Calculation module.
• Classification module by Neural Network (Deep Learning).
a. Compatible with more recent versions of NVIDIA CUDA 11.x drivers (TensorFlow 2.0).
b. Possibility of creating models trained by Geometry (in addition to the RGB and Intensity models).
c. Possibility to adjust the simplification distance and the neighborhood radius when creating trained models (important for aerial models).

‌

General
• Importing external surfaces or mesh objects with real location (ply and obj format).
• More than 2 000 ESRI projections have been added to the EPSG list
• Delete button has been moved away from add or edit buttons to avoid confusion (on all functions).
• New buttons on the toolbar to visualise Spherical images: Show, Next and Previous images.
• Enhanced speed in project creation.

Edit
• Simplify: Spherical images now follow the newly created simplified project.
• Clipping box: rotation handles problem fixed

View
• New map interface to visualize your project and spherical image locations on an OpenStreetMap, if your project is georeferenced.
• Launch spherical images visualisation directly from the map interface.
• Spherical images are now compressed when Enforced valid size is used.
• Indexing spherical images with scan numbers id.is now available with E57 project creation.
• Planar images from Leica Pegasus system can now be loaded and visualized with your point clouds.

Survey
• Fix on export Points/Lines in KML format.

Tools (new menu)
• Fix on IFC to OBJ module.
• Classification module by Neural Network (Deep Learning).
a. Performance enhanced.
b. Classification name can be transferred to VisionLidar from selected trained model.
c. Trained model now shows if it has been built with intensity or RGB.
d. Project partition is now available for single Nvidia GPU computer to enhance GPU memory usage.
e. Train a model: fix on copying all files to the correct location to be read by the classification module.
• Performance enhanced on Image to scan module.

‌

Updates
• Character "-" is again accepted in project naming when creating new projects.
• Translation/Rotation of scans now correctly apply rotation on a pivot point.
• Deep learning Classification module is now accessible with a VisionLidar Ultimate licence.
• Processing time reduced on Deep Learning Classification for VisionLidar Project in Feet.
• Fix when using Deep Learning in CPU mode

Tools
• Neural Network (Deep Learning) Classification module can now split large VisionLidar projects into partitions for better GPU memory management.

‌

General
• Database structure update augmenting efficiency and speed during re-indexation.
• New « Point Info » tool to get more information on a specific point.
• New UDP Connection protocol that allows to share last point clicked position with compatible software.
• Enhanced fence tool now allowing naming different views and store them for further use.
• Clipping box normalization for all module using it.
• Ability to save Class naming definition and load them.

Edit
• New advanced selection tool allowing selecting point clouds.
o by dragging a window,
o drawing a polygon,
o use a point radius,
o by cluster detection according by predefined criteria.

View
• Now register projects with Spherical images included (Translation applied to each image).
• Automatic detection of camera height from possible ground (Spherical Images).
• Indexing spherical images with scans numbers id.
• Now compatible with spherical images from Pegasus mobile mapping system from Leica.
• New tool to convert spherical images in other format and size.
• Now Intensity updated on the second section window.

Analysis • New tool for automatic detection of catenary wires and electric poles from selected classes. • New Clash detection tool using multiple distance and analysis shapes (catenary vs trees). • New edge detection tool with new algorithms which allows to detect edge by one click, two click (indication bearing) or multiple line (one click per edge line). • New manual cylinder creation using 4 clicks. • New Road Marking detection algorithm tool allowing the use of trajectories. • New Crash Barrier detection tool.