February 27, 2023

Exploring Lidar System Quality

Beware not all 3D Lidar Imaging Systems are created equal

DTM generated from a Ground Classfied Point Cloud

The commercial drone lidar industry is transitioning from large, heavy, and expensive to compact, light and low-cost systems. Expensive "survey grade" laser scanners made by Riegl and Teledyne are no longer the only option for commercial drone operators. Many operators mistakenly claim to use "survey grade" LiDAR systems but in fact they are using an automotive lidar scanner which is not capable of achieving "survey grade" root mean square error (RMSE) by rigorous testing. The LiDAR industry does not have a universally accepted range of RMSE for "survey grade" so it is a vague term. Land surveyors use high precision equipment such as total stations and RTK GPS that are capable of accuracies of less than 1cm and 2.5cm respectively, horizontally, and vertically. The term "survey grade" for drone lidar collected data could be reasonably accepted as achieving less than 2.5cm root mean square error.

Ultra Low-Cost DJI L1 Scanner Beam Divergence

Today, mid-grade and automotive class low-cost laser scanners are being produced by many manufacturers for use in the construction of LiDAR/Imaging systems. The lowest cost systems compromise several performance metrics to achieve the price point, such as: accuracy, precision, detectability/sensitivity, canopy penetration, beam divergence and beam quality. Ultra low-cost systems such as the DJI L1 or Rock Robotics using DJI's Livox sensors have several issues including poor precision (high noise), erroneous "air" points, and high beam divergence.

Spitfire strikes a balance by using a mid-range Lidar/Imaging system from the Geocue Group who focus on high accuracy data collection and processing for land surveyors. Our Trueview 3D Imaging system is equipped with a Hesai-Pandar laser scanner integrated with dual GeoCue photogrammetric cameras and Applanix APX-15 positioning system. It is capable of accuracies/precisions slightly above "survey grade" and performs well in vegetation and has excellent detectability/sensitivity which allows for accurate definition of narrow features such as power lines. The system includes full post-processing software that generates a ray-traced 3D colorized point cloud and geocoded images. GeoCue’s True View 3DIS product series is an inclusive hardware and software offering providing users with a complete acquisition to deliverables workflow.

Unsmoothed vs. smoothed point cloud

Processing software is key to optimizing your LiDAR/Imaging data for high quality deliverables. LP360 strip alignment fixes dynamic errors between strips to improve consistency and alignment between scan lines. The smoothing algorithm compresses the layer of 3D points while not shifting the mean which is critical to maintaining accuracy.

Unaligned vs. Aligned

To further increase the accuracy of the LiDAR data debiasing of the dataset shifts the point cloud to match the measured control points vertically when statistical thresholds are met. An effective automatic ground classification algorithm, interactive tools and sheer thoroughness are essential to isolating the last returns which estimate the ground terrain.  

Section showing Ground Classification of a Point Cloud

When selecting a commercial drone operator for a land survey consider reviewing their LiDAR/Imaging system for your project.

• What are your accuracy and precision requirements?

• Do you require accurate ground data beneath canopy?

• Do they strip align the scan lines or manually review the effectiveness of the auto ground classification algorithm?

• Do they provide quality control reports revealing LiDAR/image accuracy to ground surveyed control points?

• Does the digital terrain model produced from the LiDAR data have a vertical ground control report revealing the effectiveness of the lidar scanner in the penetration of vegetation and the effectiveness of the software and processor in ground classifying the LiDAR data?

< Back to Blog