Discover Your Island

Current Projects

1. Autonomous Searching and Tracking of Shoreline Using UAS

Background: Despite rapid technological development of UAS technology, a completely autonomous unmanned aerial system is not a reality yet, particularly for civilian or commercial uses. There are many technological hurdles to overcome before autonomous unmanned aerial systems become widespread.  Even though some low-level of autonomy has been achieved for UAS, it still requires a human operator to control the flight of an UAS. There are many applications that can benefit from autonomous UAS technology. For example, running surveillance UAS flights around the clock can be too expensive due to the need for human operators. Similarly, flying an UAS in some inaccessible or hostile location can be too inconvenient or dangerous for a human operator. It is possible though to develop in some cases an application-specific, cost-effective solution for autonomous UAS by utilizing existing results already published in literature. Accordingly, in this project, we plan to develop an autonomous UAS to search and track a shoreline. This research will achieve a technological advancement that can help researchers in many fields who collect data from such environment for their study. The system can also be used for autonomous video surveillance.  

Problem Description: A shoreline lies along the edge of an ocean that separates land from water. The separation of land from water can be detected from aerial imagery as evident from some existing works that deal with similar situations. However, there exist some additional challenges that need to be addressed in this problem due to the fact that a shoreline is very dynamic, always changing due to waves, currents, winds, and tides. We plan to investigate how to detect a shoreline using a UAS in real time so that the UAS can follow the detected shoreline as a guide to fly itself over the coastline. The research to develop autonomous unmanned aerial systems for shoreline tracking can benefit other fields of research such as environmental science and ecology. In addition, it has many other civilian, governmental, and commercial applications.

Student Research Activities:

  • Fly UAS to collect aerial imagery of the shoreline along the Gulf Coast nearby campus.
  • Implement shoreline detection algorithms and test them for effectiveness and accuracy using aerial imagery data.
  • Simulate the flight path of an UAS using results obtained from the shoreline detection algorithms to determine its feasibility for real-time applications.

Subsequently, the REU participants will develop and integrate the shoreline detection system with the navigation system and test its effectiveness in real-time environments.

Student Background:  Basic computing knowledge and introductory programming skills such Java or C++ is recommended.


2. UAS Digital Mapping for Coastal Geoinformatics

Background: Coastal management requires accurate and timely information about the state of coastal landforms (e.g., beaches) and how they are evolving. Small-unmanned aircraft systems (UAS) equipped with digital cameras provide a new paradigm for surveying of coastal environments.  Compared to traditional field surveying methods or manned aircraft surveys, UAS provides certain advantages: rapid deploy capabilities, temporal flexibility, and hyperspatial image acquisition. When coupled with repeat observations, these data can be used to derive high definition 3D digital terrain models and to characterize spatial-temporal changes of built and natural terrain in coastal zones. This newly captured information can be used to improve management practices, engineering design, and scientific decision making as it relates to coastal resiliency. For example, UAS topography data can be used to improve model calibration and forecasting of coastal hazards such as hurricane impacts or oil spill distribution.

Problem Description: We plan to utilize UAS equipped with multispectral sensors for imaging of coastal landforms within the region: beaches, dunes, wetlands. The acquired image sequences will be processed using structure-from-motion (SfM) photogrammetry to derive 3D measurements (x, y, z point cloud data) for digital modeling of coastal terrain. Accordingly, 3D data processing and spatial analysis will be implemented using geospatial computing techniques and software to characterize coastal landforms and processes captured in the repeat UAS observations. For analyzing patterns in coastal landform evolution, open-source GIS software and spatial modeling approaches, such as space-time cube (STC) representation, will be applied to the time series of hyperspatial UAS point cloud and raster products. Image classification methods using the UAS multispectral imagery will be explored to develop automated approaches for landcover mapping. Finally, through these UAS-based analyses, students will explore coastal problems of immediate relevance and importance to the region, which is an area prone to erosion and coastal hazards.

Student Research Activities:

  • Design and conduct small UAS optical imaging surveys of coastal landforms adhering to resolution and accuracy specifications as well as current regulations.
  • Process collected imagery using SfM photogrammetry workflows to derive high fidelity 3D point cloud data and orthomosaic reflectance products.
  • Assess and improve UAS-SfM data accuracy using real-time-kinematic GPS surveying methods.
  • Perform 3D point cloud data processing to generate digital elevation models (DEMs) of terrain.
  • Develop image processing and classification algorithms using UAS-acquired multispectral imagery for landcover mapping.
  • Perform GIS-based spatial analysis with the UAS data products including: extracting morphology metrics, characterizing spatial-temporal trends in terrain evolution, assessing landcover change, and deriving geovisualization products in support of coastal decision making.

Student Background: Basic computing knowledge and introductory programming skills such in C++ or Python or MATLAB is recommended.


3. Protecting the UAS against Collisions and Cyberattacks

Background: The goal of this project is to develop and test new algorithms for detecting collisions between multiple UAVs (unmanned aerial vehicles) synchronized to achieve a particular task and also to develop a robust auto-pilot system which is not vulnerable to cyber threats. UAVs were primarily used by the defense department until recently. As UAVs are now being used for non-military use, for examples, delivering pizza, delivering items as planned by Amazon, providing internet services to remote areas as planned by Google, the UAV traffic is going to increase in the coming decade. With this, possibilities of collisions and cyberattacks on them are also going to increase. UAVs can be hijacked, their paths can be changed, they can be made to collide with other UAVs or objects, and if equipped with weapons system, they can be maliciously utilized to fire in non-hostile situations. Thus it becomes critical to research into collision detection and security at the same time. In 2016, MIT students conducted a security analysis of the DJI Phantom 3 Standard and found that the drone was vulnerable to a number of malicious attacks, while researchers from Johns Hopkins University were able to use an exploit to wirelessly hack and crash a popular hobby drone. Recently on August 2, 2017, the US Military  has issued an internal memo to all its units to stop using commercial drones because of security problems. Keeping this situation in mind, robust collision detection algorithms and autopilot systems should be developed for UAVs. One of our research goals is to secure the auto-pilot system against cyberattacks.

Problem Description: When UAS are used in the field the chances of collisions cannot be ruled out and therefore, it is important to have effective collision detection strategies to avoid any collisions. The other major problem with UAS is that they are typically controlled from the ground by using a remote control which uses radio waves for communication. For a hacker it is not difficult to jam the radio waves thereby effectively making the person lose control of the UAS. Also most UAS use autopilot system which will enable the UAS to return to its base even when the controller loses access to it. This auto-pilot system is also prone to attacks by hackers. This project concentrates on examining the current autopilot systems for various vulnerabilities and cyberattacks that are common in networked systems. The major goal of this project is to design and develop a new autopilot system by using state of the art encryption algorithms keeping in mind the limited battery life.  The current security protocols available will be examined and modified to better suit the needs. In order to avoid collisions, the UAS will be equipped with cameras and sensors. By collecting the data from the cameras and sensors in real time effective collision detection and avoidance algorithms will be developed.

Student Activities:
  • Examine the current autopilot system for vulnerabilities and threats by using tools such as Wireshark, Nmap, Metasploit, Aircrack, etc. Students will be introduced to the use of such tools.
  • Analyze the current networking and security protocols available and modify them to develop new algorithms which better suit UAS as they are constrained by battery power.
  • Learn how to equip a UAS with external cameras for collision detection and avoidance.
  • Develop new algorithms for collision detection by collecting the data from the cameras and sensors in real time and then updating the waypoints of the UAS in real time.

Student Background: Basic computing knowledge and introductory programming skills such as in either Java, or C or C++, or Python is recommended.


4. Automatic Crop and Vegetable Yield Prediction on UAS Images Based on Deep Convolutional Neural Network

Background: Crop yield estimation is an important task in product management and marketing. Accurate yield prediction helps farmers to make better decision on cultivation practices, plant disease prevention, and the size of harvest labor force. The current practice of yield estimation based on the manual counting of fruits or flowers by workers is very time consuming and expensive process and it is not practical for big fields. Unmanned aerial systems provide an efficient, cost-effective, flexible, and scalable solution for product management and yield prediction. Recently huge data has been gathered from agricultural fields. However, efficient analysis of those data is still a challenging task. In this project, a deep convolutional network will be developed to facilitate the accurate yield prediction and automatic counting of fruits and vegetables on the images acquired by UAS. 

Problem description: A widely adopted solution for automatic yield estimation is to count fruits or calculate the density of flowers on images using computer vision algorithms. Computer vision approaches currently face many difficult challenges in automatic counting of fruits or flowers. In this project, we will focus on data driven object detection algorithm based on deep convolutional neural network architecture. The aim of the proposed technique is to calculate the number of ripe and unripe fruits such as tomato, apple, and orange. This research will be built upon our current research in DeepCount. The students will also work on the plant disease classification based on deep convolutional neural network which will be built on our current research on UAV image classification.

Student Research Activities:
  • Study the fundamental of machine learning and convolution neural nets.
  • Create synthetic annotated data for object counting.
  • Process the UAS data collected over AgriLife research extension in Corpus Christi.
  • Improve our deep architecture to able to count variety of unripe and ripe fruits.
  • Improve our deep classification algorithm for plant disease detection on open access data.
  • Analyze the results on both training and testing dataset and compare the results with actual ground-truth

Student Background: Basic computing knowledge and introductory programming skills in C++ or Python is recommended.


5. Exploring the Computing and Software Needs of Next-Generation Unmanned Aerial Systems (UAS)

Background: TAMUCC is one of six FAA designated UAS test sites in the nation. The primary purpose of this designation is to perform UAS research and development under FAA safety oversight. This project will allow REU students to explore the computing needs for such state-of-the-art technology.

Problem Description: Hardware-in-the-Loop (HIL) simulation is a technique that is used in the development and testing of complex real-time embedded systems. This will require investigating the diverse characteristics of the UAS such as lift, drag, time of flight, speed, payload characteristics, payload capacity, etc. in order to create a realistic HIL simulation environment, where different experimental flight setups will be tested.  As part of the data fusion system, a set of visualization tools will be developed to allow for the exploration of the data from the sensors. These tools will display the actual data collected with the ability to explore interrelationships, up-to-date values, the data series as samples, or aggregated over a specific time range. Layered and customizable maps reflecting the spatial distribution and patterns of data will also be created.

Student Research Activities:
  • Investigate a mediator-based P2P system to access underlying distributed and heterogeneous sources of marine data of variant domains.
  • Develop peer schema mapping approaches that employ coordination rules to connect two peers to facilitate queries and retrieve marine data from peers.
  • Create mediators that incorporate various query capabilities of the underlying marine data sources.
  • Test the mediator-based P2P system for reliability, consistency, and performance.

Student Background: Basic programming skills, preferably in Python and/or C# is recommended.