DJI L1 LiDAR Accuracy Revealed During Coastal Erosion Survey

• DJI L1 LiDAR achieved 15mm accuracy during coastal erosion survey - surpassing DJI's own accuracy specifications;

• M300 RTK-L1 combo took two days to collect the data: A process which would have taken three weeks using traditional methods;

• Survey team able to collect unique data which couldn't be done 'without the bigger picture from the drone LiDAR survey';  

• Highly-detailed Digital Terrain Models and Digital Elevation Models - processed using DJI Terra and TerraSolid - being used to help experts tackle coastal erosion;

• heliguy™ Drone Assistive Programme provided 'invaluable support' to DG Yeatman Surveying & Engineering Ltd;

• heliguy™ is starting a geospatial workflow course through the Learning Management System, which will include L1 best practices and post-processing in DJI Terra and TerraSolid.

The DJI L1 LiDAR sensor is capable of achieving 15mm accuracies from 100m flight altitude - a coastal erosion survey has revealed.

Incredibly, the results surpass DJI’s quoted specifications, which state the L1 can obtain 5cm (50mm) vertical accuracy from 50m.

Click to access the L1 LAS dataset

Please note: There are different ways of interpreting the accuracy of the data. The 15mm accuracy quoted in the article is based on the average accuracy difference between the ground control points and the DTM. However, the absolute average error (basing the calculation on the absolute error values) shows the L1 achieved 31mm vertical accuracy. In both methodologies, the L1 surpassed DJI's specifications of 50mm vertical accuracy from 50m.

The findings follow a mapping project at Pagham Harbour, West Sussex, conducted by heliguy™ and DG Yeatman Surveying & Engineering Ltd. This interactive orthomosaic, using data collected from the L1, shows two sections of the survey site.  

It took the L1 and M300 RTK drone just two days to collect the data over a 4.34 square KM area, with short tidal windows reducing the operating time. The image below, taken from Google Earth, shows the extent of the survey area. 

In contrast, it would have taken three weeks to collect this level of data using traditional land-based methods.

Utilising DJI Terra, TerraSolid and ArcGIS Pro post-processing software, the data was used to create highly-detailed and visual Digital Terrain Models (DTM) and Digital Elevation Models (DEM) - providing unique and data-rich insights to help monitor and mitigate coastal erosion.

This DEM, for instance, shows a large shingle spit (in the middle of the model) which is a major factor in the retreating coastline.

Darren Yeatman, Director of DG Yeatman, said: “The L1 is an impressive platform. The verified accuracy it achieved surpassed expectations and the quality of the DTM it can produce is excellent.”

DJI L1: An Accurate LiDAR Drone Solution

How accurate is a drone survey? It's an important and frequently-asked question when it comes to aerial mapping.

The surveys at Pagham demonstrate that the L1 is an extremely reliable solution.

The high levels of accuracy were verified when comparing the Digital Terrain/Elevation Models, like the DEM below...

...against manually-collected data points for the same locations, using GCPs and a Leica GNSS receiver. 

The table below shows the accuracy difference between the GCPs and the DTMs/DEMs at specific GCP locations around the site. Note how the average accuracy difference is -0.015, verifying the 15mm accuracy of the L1 for the survey at Pagham. 

However, it is also important to point out that the true average error (converting all negatives to positives in the Accuracy Difference column and basing the calculation on the absolute error values) is 0.031. This shows that the L1 achieved 31mm vertical accuracy - which is still a highly impressive return.

Reflecting on the results, Darren said: “I was very impressed with the accuracy of the L1.

“On the basis of our surveys, the system proved to be even better than the accuracies quoted as part of the product specifications.

“It’s not to say that you will achieve these results on every survey, but it does show what is possible.

“Certainly from a workflow point of view, I felt that we used the system and the post-processing software properly and conducted the right quality checks.

“As a result of this thorough and comprehensive workflow, coupled with best practices, we managed to achieve, and verify with independent checks, these high levels of accuracy.”

To help ensure the best accuracy possible, the L1 was set to the NADIR position (the point directly below the camera at ground level) and NTRIP corrections were used throughout the flights. The surveys were conducted using a mapping mission in the DJI Pilot app.

The data was verified in OSGB36 and ODN (Ordnance Datum Newlyn).

Cross-Section

The high levels of accuracy were obtained because of the quality of the L1, which returned low levels of noise during the surveys. 

This is demonstrated by the image below.  

The left view shows the end of the spit from an aerial perspective colourised with realistic RGB colour, while the right shows a cross-section of the same model coloured by elevation.

The cross-section displays a 30mm slice of the model, and displays data variance within that slice. In this scenario, this elevation view shows the minimal elevation variance in the models created from the L1, but it is worth noting that sharp elevation changes in the 30mm slice could be responsible for elevation variance rather than noise. 

Deep Insights From LiDAR Data   

The accuracy of the L1 models means that surveyors can conduct detailed measurements and make key decisions about mitigating coastal erosion - safe in the knowledge that they are working from highly-reliable and robust digital assets.

For instance, this raw RGB coloured point cloud shows a section of the site, including the large shingle spit.

The spit is pictured to the left centre of the picture.

For deeper insights, the data set can be turned into a Digital Elevation Model (DEM), as pictured below.

This model, of the same spit, shows the levels of sediment which have built up, with the red denoting the formation's highest parts.

The colour gradients, coupled with the specific elevation measurements, enable decision-makers to evaluate the problem areas, accurately gauge the height of the material, and mitigate accordingly.  

Notice, too, how sections of the spit have already been removed - illustrated by the strips of yellow within the block of red.

Incidentally, this data could be compared over a period of time to measure and monitor the spit's behaviour.

LiDAR is a particularly effective tool for building DTMs/DEMs, and the drone provides a unique aerial viewpoint. 

DJI Terra And TerraSolid Workflow

The DTMs/DEMs were produced through DJI Terra and TerraSolid.

The workflow began with uploading the raw point-cloud data to DJI Terra for initial processing...

...before uploading to TerraSolid...

...which provides extra processing capabilities to turn raw point clouds into usable data for deeper analysis and insights.

TerraSolid offers a comprehensive suit of tools to conduct classification and convert the information to the British coordinate system, also known as OSGB36.

It can also be used to clean and remove noise from a data set, as demonstrated in the next two images.

The first picture shows one of the unprocessed data sets.

It's worth noting that the raw data is pretty clean - testament to the quality of the L1 and the mapping parameters set for the surveys - but there is still a degree of noise. This is shown by the vertical cross-section on the right

However, TerraSolid can be used to manually reduce the noise through point thinning and smoothing without losing much of the accuracy and detail - as demonstrated in the second image.

Geospatial workflow drone surveying course

To help surveyors make the most of their L1 data, heliguy™ is starting a geospatial workflow course, which will include L1 best practices and post-processing in DJI Terra and TerraSolid. This course will be accessible via the online Learning Management System and will include video tutorials and PDF guides.

L1-M300 RTK: A Highly-efficient Solution

The detail of the maps and models produced for Pagham Harbour was thanks, in part, to the quality and the amount of data that the M300 RTK and L1 can collect.

Not only did the solution dramatically enhance mission efficiency, but it was used to reach otherwise inaccessible locations within the survey site.

Darren said: “For a survey of this magnitude and complexity, a drone was the ideal solution and, by using the M300 RTK and L1, we were utilising the best system available.

“Traditional land-based survey methods would have taken significantly longer and the amount of data collected would have been limited through health and safety constraints and accessibility issues.

“In contrast, the drone enabled us to access difficult-to-reach areas and LiDAR ensured accurate data collection, even in low-light environments.

“Thanks to the efficiency of the M300-L1 combination, deploying the drone was much more cost-effective. It also means that we are able to conduct regular surveys to provide a more up-to-date, real-time view of the situation. This is particularly effective when quick data collection is crucial, such as following a storm event.”

The team has also been using the M300-L1 to model the existing harbour, to demonstrate the way the harbour responds to the tidal cycle.

"This is something we would struggle to achieve without the overview obtained from the drone LiDAR survey," said Darren.

The L1 is a multi-purpose payload, capturing LiDAR data as well as housing a 20MP sensor for photogrammetry. This enables the creation of real-time true-colour point clouds, as well as detailed 2D orthomosaics, like the one below.

Darren said: “The fact that the L1 is a dual-mapping solution means that you can combine two applications at once: You can undertake both a LiDAR and aerial photography survey, thereby allowing two potential deliverables to be created from a single flight.

“It also means that you can corroborate your data, comparing your LiDAR data sets with your photogrammetry data.”

Using the photogrammetry data, the team has also been able to stitch the whole site into one composite model to provide a full and comprehensive overview in one image.  

While this orthomosaic isn't the best quality - due in part to the different shades throughout the map, as a result of varying mapping conditions/times - it does show the value of such a highly-visual asset which is fairly unique to drone data collection.

Click to access the L1 LAS dataset

Using the Heliguy Drone Assistive Programme

DG Yeatman has been supported through the heliguy™ Drone Assistive Programme: A package which nurtures the growth of in-house drone programmes through comprehensive supply and support and real-world industry-application training.

This included heliguy™’s in-house GIS team of Ben Sangster and Richard Dunlop providing hands-on assistance for the two-day survey at Pagham.

DG Yeatman has also benefited from a heliguy™ hybrid GVC training course, which included fast-track practical flight training to ensure operational capabilities, as well as a thorough M300 RTK and L1 handover day

Darren said: “The heliguy™ Drone Assistive Programme has been invaluable. It has been a tremendous help to get us started and begin to realise the true value of drone technology. We have received excellent, well informed and invaluable service from the entire heliguy™ team right from the start."