“Point cloud classification will be very useful for feature extraction. Great job so far. We really like it.”
- Alex Garcia, National Manager, Mobile Solution, GeoVerra
Rapid advances in artificial intelligence (AI) capabilities provide numerous advantages for surveyors as they integrate and streamline data management in a data-abundant ecosystem. By training machines to do time-consuming, repetitive activities, such as point cloud classification and feature extraction, humans have more time available to complete the most valuable components of the workflow.
Adding Value to Data
In the modern world, accurate raw data is not enough. The objective is to combine and transform the collected data into information that supports confident decision-making. In Trimble® Business Center (TBC) office software, AI is being used to mimic human cognitive functions with a goal of high-quality replacement, meaning that the result produced by AI-powered tools is expected to be as close as possible to the quality of human intelligence.
With the help of AI, TBC removes redundant manual interaction from the workflows and allows surveyors to concentrate on the tasks most suitable for higher-level thinking and problem-solving. The productivity gains provided by AI-powered tools are accessible to everyone without requiring AI expertise. Surveyor expertise becomes even more critical as it is applied primarily to the complex, high-value portions of the workflow.
TBC is designed to work with all types of sensors and to manage any type of data—whether it is from a drone, a terrestrial scanner or a mobile mapping system—which results in a centralized database of information to support reliable and robust deliverables.
Automated Point Cloud Classification
“The classification process is getting much easier and automated. More comments to come soon! Thanks for staying on top of the game!”
- Mark Keeton, RPLS| Jacobs |Group Leader - Survey/Geomatics, South Central Region
Point cloud classification is an essential processing step for many Trimble users. In TBC, automated point cloud classification is based on a 3D deep-learning semantic segmentation model. This advanced model is trained on representative datasets that encompass a broad range of geographic locations to ensure coverage of as many user groups as possible. 3D deep learning is more robust because it learns key characteristics on its own.
The former algorithm-based approach defined certain rules for classification, and if the object did not look familiar, the algorithms would fail. The latest AI models adapt to a variety of challenges when classifying point clouds and address these challenges for the user.
Automated point cloud classification is based on a 3D deep-learning model.
In TBC, users can apply AI to classify point clouds into buildings, vegetation (high and medium), poles, signs, ground, noise (people and vehicles), steps, power lines and dividers. This 100% automatic functionality does not require setting complicated parameters; it is as simple as selecting classes in the user interface, clicking the Classify button, and letting it run.
Considering that computers are available 24/7 and reliable algorithms avoid human errors, AI provides significant productivity improvements compared to manual classification. To try this feature within the TBC software, select Point Clouds > Extraction > Extract Classified Point Cloud Regions.
The automated system classifies people and vehicles as “noise.”
“I like the fit and finish of this command. Refinement tool is spot on.”
- Jarrod Black, Senior Vice President - Survey, Rochester DCCM.
Feature Extraction with Detailed Attributes
AI is also leveraged to improve productivity and accuracy for labor-intensive feature extraction activities. With minimal user interaction, TBC extracts the location, attributes and geometry for each individual object of interest. In addition to the time saved, the information gathered is more detailed and reliable than when the same task is performed manually.
AI is suitable for all kinds of projects to expedite feature extraction. In addition to providing the location for objects like poles, signs and trees in the point cloud, detailed parameters such as height, diameter and inclination for poles and signs and diameter, height and the crown spread for trees is provided. These tools are completely automatic and are based on a set of algorithms. To try this feature, select Point Clouds > Extract Point Feature > Extraction Type > Tree or Pole or Sign in TBC.
Detailed parameters about each object are generated in Trimble Business Center.
Manhole extraction is also accomplished with a completely automated command that delivers the location and diameter of each manhole based on a 2D deep-learning object detection model trained on the datasets from different geographic locations.
Within TBC, the point cloud is rasterized, the center of each covered manhole is located, and the diameter is measured and extracted. To try this feature, select Point Clouds > Extract Point Feature > Extraction Type > Manholes - Laser Scanner or Manholes - Photogrammetry in TBC.
Countless hours are saved by automatically extracting manholes and road markings.
TBC has dedicated tools for extraction of road curbing and gutters, and road markings of any complexity, including dashed and solid lines, crosswalks, parking spaces and more. These tools are semi-automated and are based on a set of algorithms, including template matching techniques, that recognize the objects in the point cloud.
A new workflow offered in TBC version 5.90 leverages AI to simplify and expedite extraction and volume computation of stockpiles in point clouds. By identifying and removing all unnecessary user clicks, a single command now offers a complete workflow and highly accurate results.
‘’I have been testing the workflow and I think it will get a lot of our clients very excited.’’
- Mike Tartaglia, Surveyor/Technology Specialist, SITECH Southwest
Expedite extraction and volume computation of stockpiles in point clouds.
The stockpile picker feature automatically generates the boundary around the selected stockpiles across a construction site, while the “Calculate Volume” button calculates stockpile volumes, slope and base area, and other measurements with little manual interaction necessary. This information can be exported either as a TBC report or directly to a CSV file for further analysis. To try this feature, select Point Clouds > Extraction > Extract Stockpiles in TBC.
“The volume computation automation is very clear. There are many possibilities behind this feature because stockpile volume determination is a critical workflow stage. I love the stockpile extraction—it is very useful.”
- Ariel Silva, Gerente de Soporte y Preventa (Support & Presales Manager), Geocom
Maximum Efficiency with AI
As the amount of data being collected and managed continues to increase, streamlining workflows and scaling production with computer-assisted operations is critical for reducing costs and improving quality. TBC’s 3D deep-learning semantic segmentation model automatically classifies point clouds into buildings, vegetation (high and medium), poles, signs, ground, noise (people and vehicles), steps, power lines and dividers, leaving humans more time to perform complex analysis on the results.
The automated feature extraction capabilities in TBC extract the location, attributes and geometry for each individual object of interest, including poles, signs, trees and manhole covers. The greater detail and higher accuracy obtained with AI compared to manual feature extraction supports better analysis and delivers significant time savings. The stockpile extraction command calculates stockpile volumes, slope and base area, and other measurements with little manual interaction, also saving labor hours.
By leveraging AI, TBC’s automated feature extraction and point cloud classification functionality provides actionable information to support urban planning, highway maintenance, vegetation management and many other valuable applications.
To learn more about Trimble Business Center or get a 30-day free trial, click here.