Versions Compared

Version Old Version 1 New Version 2
Changes made by

Reza Mahmoudi Kuohi

Reza Mahmoudi Kuohi

Saved on

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

Menu: AIAnalyze

Version française

VisionLidar’s AI-powered tools bring automation and efficiency to LiDAR data processing, enabling advanced classification, segmentation, and model training. This section provides access to four AI modules designed to streamline workflows and improve accuracy.

AI Modules Overview

1. AI / Automatic Classification

🔹 Automatically classify LiDAR data using pre-trained or custom AI models.
🔹 Choose between GPU or CPU processing for performance optimization.
🔹 Supports corridor-based classification for linear projects.

📖 Learn more → [Automatic Classification Guide]

2. AI / Train Model

🔹 Train a new AI classification model using LiDAR point cloud data.
🔹 Choose between default profiles or custom training parameters.
🔹 Optimize training settings with validation splits, GPU acceleration, and performance modes.

📖 Learn more → [Train Model Guide]

3. AI / Export for Train

🔹 Prepare point cloud data for AI model training.
🔹 Two export options:

Fence Selection → Select part of a classified point cloud using a fence and export it as .PLY.

Full Export → Directly export the entire point cloud as .PLY for training.
🔹 Ensures high-quality training datasets for optimal AI model performance.

📖 Learn more → [Export for Train Guide]

4. AI / Classify LAS

🔹 Classify .LAS files directly without creating a project.
🔹 Select input/output folders, define the model, and choose the unit of measure.
🔹 Supports GPU acceleration and optimized performance modes.

📖 Learn more → [/wiki/spaces/VLEN/pages/3487072343]

Optimizing AI Performance

✅ GPU Acceleration: For best performance, use an NVIDIA GPU (recommended for large datasets).
✅ Performance Modes: Choose Minimal, Balanced, or High based on your system’s GPU memory.
✅ Data Preparation: Ensure high-quality, well-labeled training data for improved AI accuracy.

By leveraging these AI tools, you can enhance automation, reduce manual effort, and improve classification precision in your LiDAR workflows. 🚀The Analyze menu in VisionLidar provides a suite of detection and vectorization algorithms to extract meaningful data from point clouds. These tools help classify, segment, and convert point cloud data into vectorized elements for further analysis and modeling.

Segmentation

VisionLidar includes automatic segmentation tools that categorize points based on their type or location. These tools help users segment different elements within the dataset.

  • Aerial – Classifies points obtained from aerial LiDAR scans.

  • Ground – Identifies and isolates ground-level points.

  • Buildings – Detects and classifies building structures within the point cloud.

  • Vegetation – Segments and classifies trees, shrubs, and other vegetation.

  • Noise – Identifies and removes unwanted outlier points from the dataset.

Cross-Section and Profile Analysis

For terrain and structural analysis, VisionLidar allows users to extract sections and profiles from the point cloud.

  • Sections – Generates cross-sections of the point cloud for engineering and design applications.

  • Profiles – Extracts longitudinal profiles along a specific path for road and infrastructure planning.

Object Detection and Vectorization

The Analyze menu includes tools to detect and vectorize objects from the point cloud, enabling conversion into CAD-ready formats.

  • Planes – Identifies and extracts planar surfaces.

  • Cylinders – Detects and models cylindrical objects, such as pipes and columns.

  • Trees – Recognizes and vectorizes trees in the dataset.

  • Object Detection – Finds predefined objects based on geometric properties.

  • Clusters – Segments the point cloud into distinct groups of points based on proximity.

Comparing and Analyzing Surfaces

  • Compare Scans – Analyzes differences between multiple scans for change detection.

  • Analyze Surface – Evaluates terrain or surface properties within the point cloud.

Feature Detection and Vectorization for Infrastructure and Roadways

VisionLidar includes specialized detection and vectorization algorithms for infrastructure and roadway elements.

  • Catenary Detection – Detects suspended cables and catenary wires and poles in power and rail networks.

  • Edge Detection – Identifies sharp edges and boundaries in structures.

  • Road Marking Detection – Extracts road markings from the point cloud for mapping and navigation.

  • Crash Barrier Detection – Identifies and models roadside barriers.

Planimetric Orthophotos