A Japanese research team recently published a paper outlining an innovative approach to surveying using Unmanned Aerial Vehicles (UAVs), commonly known as drones. The drones collect topographic and architectural data, which are used to create 3D computer-generated images.
The researchers aim to enhance the method's accuracy by fine-tuning the coordination of hardware and data. This would allow a more accurate assessment of construction sites and disaster areas. The paper was published in English in the Journal of Digital Life, an interdisciplinary online journal focused on digital technology.
On September 17, Nagasaki Prefecture made 3D data of nearly the entire prefecture available for free on its website, Open Nagasaki. This data is open for both commercial and non-commercial purposes. Open Nagasaki presents a grid map of the prefecture, allowing users to download 3D data for the area they select.
For instance, users can download the 3D data for Minamiyamate-cho in Nagasaki City, home to the Former Glover House, a World Heritage Site. They can view the 3D model from various angles and observe how the area gently slopes down toward the inland sea. This feature is not easily discernible from 2D aerial photographs.
UAVs With Laser Scanner Units
The essential component in generating this 3D data is UAVs equipped with laser scanning units. These UAVs collect an extensive set of "point" data, complete with location and color information, by projecting lasers from the sky to the earth's surface. The amalgamation of this data, 3D point cloud data, accurately replicates real-world topography and urban landscapes.
Local governments hope to leverage the 3D point cloud data to promote tourism and enhance disaster prevention efforts in both public and private sectors. The national government also expresses an interest in utilizing this data for monitoring construction projects in civil engineering and architecture.
However, ensuring the accuracy of the 3D point cloud data is vital, and researchers acknowledge that measurement errors are inevitable. In a 2018 study, Nakamura and his team identified three primary issues with previous measurement techniques.
The first problem lies in the decline in measurement precision when the UAV's flight speed is inconsistent. The UAV's flight pattern resembles painting a canvas with a brush, which means numerous errors can occur in data collection during acceleration and deceleration.
The second issue is associated with using data from satellite positioning systems (GNSS) like GPS without adjustments, leading to discrepancies in vertical directions.
The third challenge involves the degradation of point cloud data precision caused by the inclusion of data from distant objects measured at inappropriate laser incidence angles. For instance, when different routes are used to measure the same object, the data from closer and farther positions become mixed.
The Team's Solutions
To address these problems, the research team first reviewed the sensors and devices mounted on the UAV. This reduced the slight time disparities between devices and improved elevation measurement accuracy.
To resolve the issue of varying UAV flight speeds, highly accurate travel data was extracted from the GNSS receiver to determine flight conditions. This allowed data collected during acceleration and deceleration to be excluded. The 3D images were generated using only the data collected while the UAV flew in a straight light at a consistent speed, preserving accuracy.
The problem of deviations in elevation direction was tackled by establishing a "reference point for adjustment" at a flat location with nothing obstructing the sky. Then, the deviation of the point cloud data was adjusted in the direction of elevation.
For problems where data for the same object was measured by multiple routes, the team attempted to eliminate inaccurate data by dividing the entire point cloud data into specific-sized grids. Then, they extracted the data measured from the nearest location for each grid.
Using these methods, the research team conducted a demonstration at a drone training site in Osaka Prefecture. They utilized a UAV flying at a speed of 4 meters per second (13 fps) to gather data over a designated area. The team measured more than 100 to 200 points at each of the ten locations to assess the difference in elevation compared to existing data that was considered the most precise.
The data's precision was evaluated on a four-point scale from A to D, used in the guidelines for work progress management prescribed by the Ministry of Land, Infrastructure, Transport, and Tourism. Out of the points obtained using the new method, 80 to 90% were rated A (difference of less than 0.05 m) or B (difference of 0.05 m to 0.10 m).
The measured values remained within the device's margin of error. This led the researchers to conclude, "We have confirmed that the proposed method has solved the three problems of the existing method."
Various Methods of Collecting 3D Point Cloud Data
Research team member Professor Nakamura notes that various local governments throughout Japan are releasing 3D point cloud data. Shizuoka Prefecture led the way by starting this project around 2020. At the moment, these point cloud data encompass a range of methods and results. Among them are laser profiler data, which are comprehensive measurements of the entire country using lasers irradiated from aircraft and rotary-wing aircraft. MMS data are collected by road vehicles and combine photographs and data collected by lasers and GNSS to measure road spaces. Then there are data obtained by UAVs, which measure relatively small areas with high density and precision.
Why Laser Surveying?
Nakamura highlights the unique advantages of their method, emphasizing its ability to measure smaller areas at a higher density compared to laser profilers.
He explains, "Unlike UAV photogrammetry devices, laser surveying can collect data at night and in mountainous areas. This makes it possible to use this technology to supplement point cloud data that is already available. In the event of a disaster, it is also possible to measure 3D point cloud data on-site and compare it with open data to determine the extent of damage."
The research team consists of Professor Kenji Nakamura of Osaka University of Economics, Professor Shigenori Tanaka and Assistant Professor Yuhei Yamamoto of Kansai University, Professor Ryuichi Imai of Hosei University, Associate Professor Yoshinori Tsukada of Setsunan University, and Assistant Professor Masaya Nakahara of Osaka Electro-Communication University.
- How Useful is Artificial Intelligence in Traffic Surveys?
- Successful Drone Test Shows Potential in Disaster Prevention and Response
- Expert Sounds Alarm on Unknown Megathrust Earthquake in Kanto Region
- Heavy Rain Forecasts by Supercomputer System in Latest Update to Japan's Resilience Plan
Author: Taketoshi Noma