Coenradie uses AI-Driven Point Cloud Classification with Leica Cyclone 3DR

Writer: Mary Jo Wagner 

With the rapid march of reality capture seemingly reaching everything and everywhere, there has been a tremendous emphasis on creating interconnected workflows and transforming rich 3D data into meaningful, actionable information. That push to see, understand, and achieve more is redefining business as usual for geospatial users in various industries, such as surveying, architecture, construction, manufacturing, and utilities. 

A key enabler for creating meaningful analysis and modelling is efficiently and accurately cleaning and classifying the millions of data points captured by terrestrial laser scanners (TLS), mobile scanners, mobile mapping systems (MMS), and aerial systems. Advancements in artificial intelligence (AI)-based classification not only increase the efficiency and productivity of feature extraction but also enhance downstream applications and value-added deliverables.

“Manual classification, particularly with massive reality capture datasets, is often hideously time-consuming and vulnerable to errors,” says Sander Schröder, innovations manager at Coenradie, an engineering consultancy headquartered in Oirschot, The Netherlands. “AI classification significantly expedites processing and improves accuracy, which allows us to create higher quality deliverables for clients.”

Coenradie has been using point cloud software Leica Cyclone 3DR to automate, improve accuracy and speed up their point cloud classification workflow. This move meets their need for speed and brings efficiency and ease to their downstream applications, enabling them to pursue new opportunities.

Cyclone 3DR has become the defacto solution for creating meshes, models, drawings, and reports for Coenradie, a veteran and disruptor of survey solutions.

Capitalising on classification models

As a 30-year veteran providing survey solutions, Coenradie has been driven by one primary goal: to be “the first to the future.” As such, the company has routinely embraced cutting-edge hardware and software to help redefine what’s possible for its industrial, governmental, architecture, shipbuilding and construction clients.

As an early adopter and industry disruptor, Coenradie first began using Cyclone 3DR, an all-in-one deliverable and analysis software solution, to improve its point cloud analysis and modelling workflows and create detailed digital reality deliverables to enhance their client's decision-making.

“Cyclone 3DR has become our de facto solution for creating meshes, models, drawings, and reports,” says Schröder. “It automates common tasks and reporting, which frees us up for more complicated tasks; it’s saved us 25 percent on registration time and yielded up to 70 percent savings on calculating volumes for large projects. So, it’s made a significant difference in our efficiency.”

Now, Coenradie is capitalising on the software’s seven pre-defined classification models, powered by AI, to automatically segment, clean and filter TLS, MMS, mobile scanning and UAV data. The AI classification is not only transforming feature extraction into a minutes-long exercise; it’s smartening their 3D data deliverables and helping them re-engineer traditional field-to-office workflows.

Automatically clean and classify objects in a point cloud with one click when using Cyclone 3DR AI Classification.

Clean, classify, clarify: faster, smarter point cloud cleanup with Cyclone 3DR

With each new AI classification model release, Coenradie has been able to streamline its data processing and modelling workflows, improve the quality and flexibility of its deliverables and enhance the integration of 3D datasets into its clients’ work routines. 

“If I’m asked to describe the process involved in cleaning and classifying reality capture data, it’s almost a bit embarrassing to answer because it’s so simple,” says Jan Barnas, a Coenradie 3D specialist. “It’s literally just pick a classification profile and click ‘go’.”

Unlike in the past, where classifying complex piping components often required experts to manually interpret and label elements that were either missed or misclassified, Cyclone 3DR’s industrial classification model only requires the click of a mouse. The software automatically cleans the point cloud and uses pre-defined classes to categorise each pipe component, easing Scan-to-BIM processes and creating more intelligent, customised models.

“What really impresses us is how quickly it cleans the point cloud and leaves just the piping,” says Barnas. “Its ability to accurately classify the features we need and filter layers gives us so much flexibility for clients. We can produce drawings of only what the customer wants, or we can split the point cloud up and deliver only the sections the client needs. That’s just gold.”

Classifying and filtering general indoor environments like office buildings has also benefitted Scan-to-BIM applications.

“If you have a five-story building filled with furniture and you need to filter all that out, it would be incredibly time-consuming to go to each floor, identify each chair or table and manually create a bounding box around each piece to eliminate it,” says Schröder. “With Cyclone 3DR classification, you press a button, wait a few hours, and it’s done. That dramatically speeds up our modelling and opens up tremendous opportunities for us and our clients.”

A new approach to modelling urban environments 

Coenradie is also experiencing the same speed, precision and customisation in classifying outdoor areas for modelling urban environments from topographic surveys.

Classifying trees for efficient point cloud cleaning and registration is a daily challenge. AI classification’s General Outdoor Model enables Coenradie 3D data specialists to automatically identify and filter out trees to geometrically register point clouds accurately. For clients who want to produce their own CAD drawings, they can automatically generate a clean cloud devoid of trees and offer the option to turn the tree layer on if they are relevant to the design.

Traditionally, Coenradie has used total stations and GNSS technology for topographic surveys, but the advancements in Cyclone 3DR’s AI classification and mobile scanning have enabled them to transform their modelling workflow from the field to the office. This approach first served them on an urban planning project for the town of Deventer in 2023.

In about 30 minutes, Cyclone 3DR's AI classification feature categorises all objects of interest including buildings, curbs, tree types, parking spots, and utility poles.

Classified intelligence for urban planning

With a focus on planting more trees and creating more pedestrianised and cycling pathways in Deventer’s Niewe Markt area, urban planning company StadLandWater Management en Advies tasked Coenradie with acquiring and producing a point cloud and CAD drawings of a 250m x 250m area.

The team capture point cloud data in the field using 3D laser scanners such as the Leica RTC360 and Leica P50 laser scanners. Once the data has been collected, Coenradie ingest the scan data and panoramic imagery into Cyclone 3DR for processing. Running both the Outdoor Environment and Road Classification models, the software systematically cleaned the 3D data, freeing it of any clutter like birds, cars or people, and removed noise caused by shiny objects. The remaining objects, such as trees, were classified based on each model’s separate classes. In about 30 minutes, Cyclone 3DR’s AI classification feature had categorised all the objects of interest, such as buildings, tree types, curbs, road centerlines, parking spots and utility poles.

“Having to classify all those features manually would have taken a full day’s work, if not longer,” says Barnas. “With Cyclone 3DR’s classification, we were ready to begin our next steps in less than one hour, which was 90 percent accurate. That was a huge time saver.”

With the classified point cloud ready, Coenradie 3D Specialists extracted both a CAD drawing and digital PDF report containing detailed scene data including tree locations and types, road/pavement height in specific spots, parking spaces, curbs, pavement and pavement type, road centerlines, and light pole and waste bin locations. They also provided an online viewer for the client to see the point cloud and panoramic images online, enabling them to measure distances, angles, and surface areas easily.

“We host the data online, so we know they look at the detailed point cloud nearly every day,” says Barnas. “It’s another way of providing additional value to them.”

"Cyclone 3DR will enable us to continue to innovate, transform traditional tasks and push the boundaries of what new realities we can create.”

Shaping the future with classification

Anchored by Cyclone 3DR’s AI classification’s efficiency, flexible oversight, and intelligence, Coenradie sees endless opportunities to expand their reality capture solutions.

“Leica Geosystems classification really is the basis for making everything easier downstream in terms of further data analysis or production of other deliverables,” says Barnas. “It will enable us to continue to innovate, transform traditional tasks and push the boundaries of what new realities we can create.”

That kind of pioneering spirit will undoubtedly continue to support Coenradie in its quest to be the first to the future. 

Interested in learning more about AI Classification and other features and workflows in Cyclone 3DR? Browse our Cyclone 3DR YouTube Channel for tutorials, tips and tricks. 

Yannick Stenger
Product Manager, Leica Cyclone 3DR