UNDERWATER OBJECT DETECTION USING YOLOV5
Introduction
Underwater wireless networks enable monitoring, surveillance
applications
Computer vision important for automated analysis of underwater
data
Object detection with deep learning can process large volumes
quickly
YOLOv5 Model
Directly predicts bounding boxes and class probabilities
Simpler and faster compared to classification-based approaches
Trained on underwater images captured using Raspberry Pi camera
Results
80 images with augumentation used for training and testing
Achieved 75% accuracy and recall, 85% F1 score for 1-class model
Precision fluctuates more due to limited dataset size
Predicted bounding boxes match objects in test images
Analysis
Accuracy - fraction of correct predictions
Precision - percentage of correct positive predictions
Recall - finding all positive cases
F1 score - combines precision and recall
Conclusion
YOLOv5 provides decent underwater object detection
More data and classes will improve training and accuracy
Promising for automated analysis of underwater visual data
Key Highlights
Deep learning enables efficient computer vision for underwater
networks
YOLOv5 model straightforward for custom object detection
Performance metrics quantify accuracy on test data
With more data, can enable underwater monitoring at scale
SECURITY IN INTERNET OF THINGS (IOT)
Introduction
IoT connects physical objects to the internet for monitoring and
control
Security critical due to growing connectivity and cyber risks
IoT architecture has perception, network, processing and application
layers
Security Threats
Physical, software, cryptanalysis, environment-based attacks
Scalability, connectivity, end-to-end security challenges
Authentication, identity protection also issues
Attacks on IoT
Happen across data collection, storage, processing, transmission,
delivery phases
Target different architecture layers like network, perception, services
Sinkhole, selective routing, witch attack compromises perception
layer
Security Enhancements
Strong authentication and access control mechanisms
Encryption for data security during transit and storage
Regular software updates to patch vulnerabilities
Physical tamper-resistant enclosures for devices
Conclusion
Multi-layered defense needed spanning hardware, software, network
Security must cover architecture components and data lifecycle
stages
Combination of prevention, detection and response required
Key Highlights
Rapid IoT growth escalating security risks
Attacks exploit vulnerabilities across device-cloud continuum
Holistic solutions integrating encryption, access control, physical
security needed
Security an ongoing challenge requiring continuous enhancement
DESIGN AND PERFORMANCE OF A COMPACT ULTRA-WIDEBAND
ANTENNA FOR WIRELESS APPLICATIONS
Introduction
Rapid advancement of wireless systems requires compact,
integrated, multiband/broadband antennas.
Nature-inspired designs like biomimetics help achieve this through
integration with engineering.
IEEE802.15 uses 3-10 GHz spectrum for high data rate UWB
applications.
FCC allocated 7.5 GHz bandwidth from 3.1-10.6 GHz for unlicensed
UWB use.
Antenna Design
Bio-inspired antennasAcan be divided into animal and plant
categories.
Proposed antenna evolved from semicircular patch by cutting
sinusoidal petals.
Final design isAlotus-shaped patchAwithAstaircase-shaped
defected ground.
Lotus shape with 8 tapered petals on FR-4 substrate.
Partial ground modified to staircase DGS with slots to improve
bandwidth.
Results and Discussion
Ultra-wide bandwidth fromA3.5-15 GHzAachieved.
Resonant frequencies ofA4.7 and 13.31 GHz.
Gain varies 0.21-1.8 dBi.
Radiation efficiency above 60%, maxA72.2%Aat 10.3 GHz.
Surface current at petal edges and ground plane staircase.
Bidirectional radiation patterns.
Conclusion
Novel compact 17x14 mm lotus antenna proposed.
Lotus shape with sinusoidal petals and defected ground.
Ultra-wide 11.5 GHz bandwidth from 3.5-15 GHz.
Suitable for wireless applications due to size, shape and bandwidth.
