Committed to ongoing professional development and lifelong learning, she has completed Post graduation in Network engineering and graduation in Computer Science and engineering. During her post-graduation she have done few research-oriented projects, presented papers in various International conferences and published articles in various International journals in big data analytics. In achieving this she has shown to be self-motivated, committed and determined in achieving her goals. She has also demonstrated negotiating and organizing skills, a firm sense of responsibility and capacity to work hard under pressure. as proven by her varied work experiences.

Edge Artificial Intelligence For Marine Monitoring

2021 to 2025

Project Description

Marine data is typically acquired by a field team, and then data analysis and knowledge production are conducted over the subsequent weeks and months, often using high performance computing equipment. The goal of this project is to investigate how initial processing of large datasets can be completed in near real time in the field to allow science missions to be adaptively changed, and information more readily provided to stakeholders. This results in reduced overheads in data management & handling at the system level. The project will be achieved using ‘Edge Artificial Intelligence’ (Edge AI), where AI algorithms are executed locally on an embedded computing device close to the source of the data.

Project Importance

The purpose of this project is to develop the hardware, firmware and software capabilities to apply a subset of the RDP capabilities to data as soon as it is acquired in the field, when access to the cloud is limited and when high performance computing hardware is not available. Platforms and sensors currently under development at AIMS include autonomous and semi-autonomous underwater, surface, and aerial vehicles, as well as new camera and sensor payloads to be deployed from tenders and by using divers. These all gather vast quantities of data, in particular image data. Developing the means to pre-process and interpret these data in near real-time while still in the field would enable informed decision making on cruises, thereby increasing the quality and quantity of information gathered.
These outcomes will lead to efficiency gains which will contribute to achieving the AIMS target EC2: Employ technology to double our yearly information output at half the unit cost in half the time.

Project Methods

The whole concept can be implemented using a 5-step mechanism as shown below:
Step 1: Data Collection
The first step begins by collecting training data from IoT devices and sensors and storing them in the cloud.

Step 2: Data Preparation
The next step involves data clean up, data reformatting, or pre-processing to inject additional information there by helping machine learning algorithm to find correlations between data.

Step 3: Machine Learning Model Implementation
In this step, prepared data from the previous step can be used to run various machine learning algorithms to train models. This step is crucial as it helps to compare results of various algorithms with one another and based on those results, a hybrid algorithm can be implemented for implementing a better model.

Step 4: Model Deployment
This step involves wrapping the model into a web service app that can be fed with data using REST calls and return analysis results. The web service app is then packaged into a docker container, which in turn can be deployed as an IoT Edge module.

Step 5: Model Maintenance
In the final step continue collecting data and periodically upload that data to the cloud. This data can be used to retrain and refine model, which can be redeployed to IoT Edge

Project Results

The goal of this project is to investigate how initial processing of large datasets can be completed in near real time in the field to allow science missions to be adaptively changed, and information more readily provided to stakeholders. By conducting processing closer to data acquisition, there are reduced overheads in data management & handling at the system level. This results in quicker decisions and streamlined processing. Specific outcomes of this project may include: -
• Quick-look image processing workflows or identification of unusual features whilst conducting survey missions. For example, the images acquired from a towed video system or by an Autonomous Underwater Vehicle could be instantly analysed for the presence of Crown-of-thorns starfish (COTS), with geo-referenced positions of individual COTS provided. The cruise leader may then choose to deploy a dive team for closer investigation, or in the case of external COTS control projects, for culling.
• Nested surveys using complementary platforms and sensors. For example, a drone may fly a preliminary mission and gather a photomosaic of a large area. This map could then be automatically classified with habitats of interest identified. A second more detailed mission could then be automatically generated to be conducted with an autonomous surface vehicle or autonomous underwater vehicle.
• Quality control and assurance. If the data can be pre-processed and analysed as soon as it is collected, then issues can be identified and corrected while still in the field. For example, acquired images could be analysed in real time to check that they are of suitable quality, that the light conditions are suitable, and that the camera position and movement speed is appropriate.

Keywords

Communication / Education,
Coral reefs,
Corals,
Crown of Thorn Starfish,
Marine planning,
Monitoring,
Oceanography