Enhancing Fishery Pond Security with Machine Learning Techniques
Abstract
In isolated or unmanned lake scenarios, where traditional monitoring systems often break down, fishery theft is a major threat to the aquaculture industry. Through the application of machine learning techniques for smart monitoring and anomaly detection, this study suggests an innovative approach to safeguarding fisheries lakes from theft. The proposed solution employs trained machine learning models to classify suspicious activity, identify illegal intrusions, and patrol lake borders by integrating computer vision with real-time sensor input. A variety of classifiers, including Support Vector Machines (SVM), Random Forest, and Convolutional Neural Networks (CNNs), were tried and tested for accuracy and reliability in detecting boat and human movement using a specially designed dataset derived from surveillance footage. Under different illumination and weather conditions, the system demonstrated more than 92% accuracy in recognizing potential theft situations. The experiments demonstrate that machine learning offers an economical and scalable method for enhancing security in off-grid fishing environments when integrated with low-cost hardware and Internet of Things infrastructure. Model deployment by utilizing edge-based models and connectivity with automatic alarm systems for instantaneous action are just a few instances of the present future work.
BibTeX
@inproceedings{Jayanthi2025Enhancing,
author = {Jayanthi, K. B. and Bharath Kumar, V. and Guhan, R. and Janani, A.},
booktitle = {2025 Global Conference in Emerging Technology (GINOTECH)},
title = {Enhancing Fishery Pond Security with Machine Learning Techniques},
year = {2025},
volume = {},
number = {},
pages = {1-6},
abstract = {In isolated or unmanned lake scenarios, where traditional monitoring systems often break down, fishery theft is a major threat to the aquaculture industry. Through the application of machine learning techniques for smart monitoring and anomaly detection, this study suggests an innovative approach to safeguarding fisheries lakes from theft. The proposed solution employs trained machine learning models to classify suspicious activity, identify illegal intrusions, and patrol lake borders by integrating computer vision with real-time sensor input. A variety of classifiers, including Support Vector Machines (SVM), Random Forest, and Convolutional Neural Networks (CNNs), were tried and tested for accuracy and reliability in detecting boat and human movement using a specially designed dataset derived from surveillance footage. Under different illumination and weather conditions, the system demonstrated more than 92% accuracy in recognizing potential theft situations. The experiments demonstrate that machine learning offers an economical and scalable method for enhancing security in off-grid fishing environments when integrated with low-cost hardware and Internet of Things infrastructure. Model deployment by utilizing edge-based models and connectivity with automatic alarm systems for instantaneous action are just a few instances of the present future work.},
keywords = {Support vector machines;Accuracy;Biological system modeling;Surveillance;Lakes;Fisheries;Reliability engineering;Real-time systems;Convolutional neural networks;Anomaly detection;Fishery Security;Machine Learning;Intrusion Detection;Convolutional Neural Networks (CNN);Intelligent Surveillance;Object Detection;Aquaculture Monitoring;Anomaly Detection;Real-Time Monitoring},
doi = {10.1109/GINOTECH63460.2025.11077015},
issn = {},
month = {May},
}