Study finds that certain confined spaces prohibit the use of large aerial drones to inspect bridges; areas for improvement remain to reduce the training and skill required to operate these devices.

MnDOT assessment of drone use as part of a risk- based approach to bridge inspection realizes safety, cost and quality improvements.

Date Posted

Utilizing UAS and Reality Modeling to Improve Bridge Inspection Result

Summary Information

The Minnesota Department of Transportation (MnDOT) is utilizing drones to collect and process large amounts of data during bridge inspections with the goal of improving the quality of bridge inspections and improving safety for both inspectors and the traveling public.

Inspection specific drone technology is maturing, and several models now exist that serve the inspection and asset management industry. These drones include features important to bridge inspection such as sense and avoid, infrared imaging, autonomous flights, and collision-tolerant features. When processed with modeling software, high-quality images can be collected and processed into high-quality inspection data such as point clouds, 3D photologs, and orthoplane images that can be processed by software into multiple formats that can easily be shared via the Cloud.

For restrictive access locations, collision-tolerant UAS (Unmanned Aircraft Systems) are the most cost-effective option. The study identified that there are many areas within bridge inspection that are prohibitive for imaging using a larger mapping UAS. Examples of these types of restricted access locations are:

  • Interior of tub girders, steel pier caps, and hollow abutments
  • Culverts, pipes, or tunnels with or without water present
  • Bridge deck soffit of large span bridges over water or heavily trafficked routes
  • Web faces and top flanges of large span bridges over water or heavily trafficked routes
  • High wall abutments; top of pier caps; bearings; vaulted span

For these identified areas, the concept of a collision-tolerant UAS was selected for trial. Other options were considered to access these hard to reach areas with a smaller micro-UAS, larger propeller shrouds, or additional acoustic anti-impact sensors, however the collision-tolerant UAS was the most cost-effective option to move forward with.

Suspect deficiencies or areas should be followed up with a hands-on inspection. Image sensors and processing software have become sophisticated and it is difficult to rely on camera specifications alone. The study demonstrated that focus should be on the inspector’s qualifications and their ability to determine on a case by case basis if the image quality is enough to determine with certainty the structural condition of the bridge element that is being inspected. In many cases, suspect deficiencies or areas should be followed up with a hands-on inspection. This is especially true when an inspector feels the need to use tools such as a hammer for sounding.

Inspectors are encouraged to use terrestrial photographs (photographs taken from the ground) in conjunction with aerial photographs to create high-resolution images and models. The task of capturing photographs from the ground has been routine for bridge inspection and continues to be necessary; in conjunction with aerial photographs, they can provide a more comprehensive model as the final deliverable. In several cases, terrestrial photographs alone were used to obtain high-resolution photographs and 3D models for post-inspection reviewing. For several of the studies’ structures, imagery was taken from a point-and-shoot digital camera or action type camera. The method of taking photographs from these types of platforms is similar to aerial photography, in that the user needs to assure there is plenty of overlap as a series of photographs are taken. The processing software was found to work best when images were taken in a smooth continuous path with approximately 75 percent overlap.

Models should include scales and/or ground control points (GCP) to ensure accuracy in the measuring of defects. Using GPS, a drone will place the model into the correct global position within several feet which is generally good enough for inspection purposes. The use of ground control points can add absolute accuracy and place the inspection model in the exact global position. Typically, these would be set by a land surveyor as aerial targets and their exact position would be recorded. These coordinates can be included in the model so when processed, the absolute accuracy can be set to as low as in millimeters. The accuracy depends not only on the precision of the actual GCP’s, but also on the height at which the drone is flown. The lower the drone flies, the higher image resolution is.

Cloud-based platforms should be used to clearly communicate inspection results. Performing bridge inspections with drones typically generates large amounts of data which can be difficult to share with bridge owners and decision makers, especially via email and ftp sites where company firewalls may exist. Using a cloud interface, a bridge engineer or owner can view the inspection data in 3D models without having to download and store large amounts of data. As technology has improved, the focus has shifted from the hardware to the data. Hardware is still important, and opportunities exist for improvement, especially in regard to different payload items such as non-destructive testing equipment.

The imaging field of view must face vertically and the device should have the ability to fly without the need for a GPS signal. The study identified two key features critical to UASs used in bridge inspection. The first was the imaging field of view needed to face vertically. This allows for inspection of members above the UAS such as deck soffit and interior beams. The second feature was the ability to fly without the need for a GPS signal. This is important when operating under a structure or in confined spaces. The quality of build, imaging payload, and flight software should be the best in the industry for proper safe inspection of bridge elements.