Droning On Hero Image

Droning on

It’s a bird! It’s a plane! It’s a drone! Here’s how CSU faculty are putting these aircrafts to work.


We’ve come to recognize the distinctive, motorized zzzzzzzzzz of a drone hovering overhead. Whether it’s a hobbyist capturing video with a camera-equipped drone or a proposed corporate plan for drone package delivery, these commercial unmanned aerial vehicles (UAV) seem to be everywhere.

It comes as no surprise then that drones have made their way into classrooms and research at the CSU. Yet as faculty and students work with UAVs, they remain mindful of the regulations around their use implemented by both the Federal Aviation Administration (FAA) and their respective campus.

Here are a few ways they aim to improve drone technology and incorporate it into their work.

​Ai yells "shark"

Since receiving approval from the FAA and California State University, Long Beach eight years ago, CSULB's Shark Lab has been using commercial drones to shoot video of the local coastline. Video surveys allow the team to spot white sharks and count people.

But the lab didn't have the bandwidth to process the thousands of hours of video it had captured. To find a solution, the Shark Lab partnered with Computer Science Professor Franz Kurfess, Ph.D., at California Polytechnic State University, San Luis Obispo to develop a machine-learning algorithm that could scrub the video and identify sharks and other wildlife as well as people, including differentiating between surfers, paddleboarders, swimmers and others.

“That technology now takes the hours and hours of flight video we've recorded—that would take 10 times more time for a human to go through—and can pump that through in minutes," says Chris Lowe, Ph.D., marine biology professor and Shark Lab director.

Researchers in the CSULB Shark Lab fly a drone to conduct a coastal survey for sharks. 
A screen displays a shark spotted using the drone. 

Cal Poly San Luis Obispo students are working on the program as part of class projects, senior projects and master's theses to improve and expand its capabilities. Currently, it can identify sharks and humans in still images. The goal is to allow the program to trace the movement of a specific shark throughout a video segment, which will allow the Shark Lab to better understand both shark and human behavior along the shore.

“While it's beyond the scope of what we can do here, drones like the ones we're building could be used to alert lifeguards and communities about the presence of sharks in areas frequented by humans," Dr. Kurfess says. “We don't have the capacity to actually do this on a large scale, but we hope to demonstrate that the technology to do this is available, and it is feasible."

While the collaboration is giving both biology and computer science students experience using drones and working across different scientific fields, the research is also proving beneficial in other areas.

For example, Kurfess and other faculty at Cal Poly San Luis Obispo are utilizing drone footage and machine-learning programs for operations like assessing fire damage or expanding a Continuous Forest Inventory at Cal Poly's Swanton Pacific Ranch.

In addition, the city of Seal Beach hopes to harness the Shark Lab's collected data to study how many people are going to the beach and how they are using it. “How close are they to a parking lot, or how close are they to a lifeguard stand?" Dr. Lowe says. “Now that the technology is developed, there are other applications of that technology that are valuable to the beach cities and can help them with their spatial planning."

Agricultural checkup

CSU faculty have also found a range of industry applications for drones outfitted with cameras. California State University, Fresno's Gregory Kriehn, Ph.D., professor of electrical and computer engineering, has dedicated most of his research to improving drone technology and applying it to precision agriculture.

First, his efforts are working to develop an algorithm that will automatically enable a drone's camera to stably focus on a specific point on the ground. This would reduce the effects of human error and help the camera focus on that point despite the drone vibrating, turning or moving through the air.

“Getting the system to actually point at a given location, and then to hold that location, is a critical factor," Dr. Kriehn says. “We want to try and do that on an automated basis. So, the problem is taken care of, and you can spend your time looking at the data and analyzing the pictures as opposed to trying to keep the picture in frame."

This capability would ultimately allow the drone cameras to take more accurate photos more quickly, as the pilot wouldn't need to consistently readjust the UAV or camera manually.

A drone in Dr. Gregory Kriehn’s lab. 
Fresno State students and faculty fly a drone through an orchard. 
Dr. Gregory Kriehn works on a drone. 

Photos taken from a drone are then used for Kriehn's second area of focus, in which he uses a mapping technique to create images of geographic spaces. Drones are flown over the space while they capture hundreds of photos of the ground, which are then stitched together into a complete photo of the area. Software is then used to analyze the image.

Kriehn particularly focuses on maps of agricultural fields that can then be analyzed for crop health. Infrared cameras can detect light not visible to humans, demonstrating areas where photosynthesis is or is not occurring—an indication of how healthy those plants are. The software can detect those areas for the researcher.

“You want to try and detect stress within your plants and trees before they are showing physical signs, because by the time you can actually see it, typically damage is already on its way," Kriehn says. “So, if you can start to detect the stress before it's visible to the naked eye, then you can implement mitigation strategies in order to try and correct the problem and maintain as high a yield as possible out of your crop."

The greatest benefit: time savings.

“You're saving manpower because you don't need people to go out and take a look at it," he says. “Then by getting updated snapshots on a daily or weekly basis, you can also look at trends in various parts of your field as well."

Students take flight

A drone-assisted surveying class in Cal Poly San Luis Obispo's BioResource & Agricultural Engineering Department (BRAE)—taught by Lewis Soloff, lecturer and licensed land surveyor—aims to ensure students have the drone skills needed for their future careers.

“The kind of information and the rapidity you get [from drone work] is just startling," Soloff says. “There are these revolutions in surveying technology that take place every 10 or 15 years … when a new technology emerges and a job that would take a hundred man-hours, all of a sudden takes seven. And I think that drone technology is the current case."

The course was first held in spring 2021 and will be offered for a second time this spring. It includes four focus areas, the first of which is drone-related academic material such as the history of drones and their place in the general field of aviation. The second is preparation for the FAA's Part 107 exam for a commercial drone pilot certificate. During the class last spring, a number of students earned their certificate.

“Entering the job market with your Part 107 is a real plus because … having that certificate, almost no matter what you do in the area of geographic information systems, surveying geomatics, natural resources or agriculture, is going to look good to a potential employer on a practical level," Soloff says.

Students get ready for a drone flight to map Cal Poly’s Student Experimental Farm. 
An aerial view of the Student Experimental Farm captured by a drone. 
A landscape view near the Student Experimental Farm. 

Third, the course covers drone flight skills. Students first learn to fly small drones indoors before heading outdoors to fly larger drones equipped with production cameras.

Lastly, it covers the applications of drones, and the 2021 final project required students to map Cal Poly’s Student Experimental Farm using a similar method as Kriehn. By also placing flight crosses, spots that have been surveyed on the ground and identified with a specific latitude-longitude, the photo can be placed on its actual GPS coordinates within a map.

“This is a half-day project for a survey crew to get GPS locations on five or six control points, and to fly the drone, that's another maybe 20 minutes,” Soloff explains. “For that amount of time, for less than half a day for one person, you can get all this field information that you can process yourself or send out to be processed.”

While the class emphasizes mapping, students also found ways to adapt their learnings and the analysis software to their study areas—such as monitoring plant health in agriculture or analyzing forested areas, like tree survival after a fire, in natural resource management.​

Room for improvement

No matter the purpose for which the drone is put to use, there is always a need for computing resources if the drone is to operate autonomously—but relying on ground-based resources, like a laptop, can result in transmission delays and requires communication infrastructure. Junfei Xie, Ph.D., assistant professor of electrical and computer engineering at San Diego State University, is working to solve that dilemma and expand the situations in which drones can be used, thanks to a National Science Foundation grant.

“The grant will allow me to develop a theoretical framework to enable networked airborne computing—a new computing paradigm formed by aerial vehicles equipped with computing resources that can talk to each other by direct flight-to-flight communication links," she says. “It can be considered a flying cloud computing system formed by computing servers that can move autonomously in the air."

In essence, the computing system and the communication device would be located within the drone itself, rather than in a remote location. This would significantly reduce transmission delays, as all computations can be performed directly onboard the drone.

“We are able to improve the computing power of the drone, and with that, we can allow the drones to implement advanced algorithms to improve their own performance," Dr. Xie explains. “For example, safety is very important for the drones, and we want them to be able to quickly respond to any obstacles and not collide with other drones, birds or even people. With powerful onboard computing capabilities, we can implement advanced collision avoidance algorithms in the drone to achieve this."

Dr. Junfei Xie holds a drone. 
Two drones in flight. 

The drones would also be able to share resources and cooperate with each other through their in-flight communication connection. With this capability, when multiple linked drones are deployed simultaneously for specific uses, like agriculture, land surveying or natural resources management, drones with high computing power will be able to accomplish the task more quickly, accurately and safely.

But it also means these drones can provide communication and computing services in regions where the communication infrastructure does not exist, such as disaster zones or rural areas. “People in the disaster areas may not be able to communicate with people outside, and they need to send messages to their family or emergency services,” Xie says. “You can fly the drones there and then provide the services they need.

“Unmanned aerial vehicles, or drones, have many unique features, like they’re low in cost, can be easily maneuvered and can be quickly deployed,” she continues. “These unique features of the drone then make the networked airborne computing system have many advantages, compared with the traditional ground-based computing systems.


​F​or more, read about a summer program to prepare high school students for the Part 107 exam at California State Polytechnic University, Humboldt.

 
2/21/2022