An Introduction to Deep Learning‐Based Object Recognition and Tracking for Enabling Defense Applications DOI

Nitish Mahajan,

Aditi Chauhan,

Monika Kajal

et al.

Published: Feb. 7, 2024

Object monitoring and surveillance technologies are crucial in defense, border protection, counter-terrorism operations. These enable military security personnel to monitor track the movement of objects individuals high-risk areas, detect potential threats, respond effectively intrusions or attacks. In object used see troop movements, enemy activities, provide real-time intelligence commanders. include radar systems, unmanned aerial vehicles (UAVs), satellite imagery. By providing early warning movements these help quickly effectively, increasing their chances success. illegal crossings, drug trafficking, smuggling activities. thermal imaging cameras, ground sensors, UAVs. information about control apprehend reducing risk incursions other threats. operations, threats prevent terrorist facial recognition biometric scanners, advanced systems. identifying dangers before they can carry out attacks, activities safeguard public. conclusion, critical enabling national protect citizens from harm.

Language: Английский

GeoAI for Large-Scale Image Analysis and Machine Vision: Recent Progress of Artificial Intelligence in Geography DOI Creative Commons
Wenwen Li, Chia-Yu Hsu

ISPRS International Journal of Geo-Information, Journal Year: 2022, Volume and Issue: 11(7), P. 385 - 385

Published: July 11, 2022

GeoAI, or geospatial artificial intelligence, has become a trending topic and the frontier for spatial analytics in Geography. Although much progress been made exploring integration of AI Geography, there is yet no clear definition its scope research, broad discussion how it enables new ways problem solving across social environmental sciences. This paper provides comprehensive overview GeoAI research used large-scale image analysis, methodological foundation, most recent applications, comparative advantages over traditional methods. We organize this review according to different kinds structured data, including satellite drone images, street views, geo-scientific as well their applications variety analysis machine vision tasks. While tend use diverse types data models, we summarized six major strengths (1) enablement analytics; (2) automation; (3) high accuracy; (4) sensitivity detecting subtle changes; (5) tolerance noise data; (6) rapid technological advancement. As remains rapidly evolving field, also describe current knowledge gaps discuss future directions.

Language: Английский

Citations

64
An Adaptive Attention Fusion Mechanism Convolutional Network for Object Detection in Remote Sensing Images DOI Creative Commons
Yuanxin Ye,

Xiaoyue Ren,

Bai Zhu

et al.

Remote Sensing, Journal Year: 2022, Volume and Issue: 14(3), P. 516 - 516

Published: Jan. 21, 2022

For remote sensing object detection, fusing the optimal feature information automatically and overcoming sensitivity to adapt multi-scale objects remains a significant challenge for existing convolutional neural networks. Given this, we develop network model with an adaptive attention fusion mechanism (AAFM). The is proposed based on backbone of EfficientDet. Firstly, according characteristics distribution in datasets, stitcher applied make one image containing various scales. Such process can effectively balance proportion handle scale-variable properties. In addition, inspired by channel attention, spatial also introduced construction mechanism. this mechanism, semantic different maps obtained via convolution pooling operations. Then, parallel are fused proportions factors get further representative information. Finally, Complete Intersection over Union (CIoU) loss used bounding box better cover ground truth. experimental results optical dataset DIOR demonstrate that, compared state-of-the-art detectors such as Single Shot multibox Detector (SSD), You Only Look Once (YOLO) v4, EfficientDet, module improves accuracy has stronger robustness.

Language: Английский

Citations

58
A multitask model for realtime fish detection and segmentation based on YOLOv5 DOI Creative Commons

QinLi Liu,

Xinyao Gong,

Jiao Li

et al.

PeerJ Computer Science, Journal Year: 2023, Volume and Issue: 9, P. e1262 - e1262

Published: March 10, 2023

The accuracy of fish farming and real-time monitoring are essential to the development "intelligent" farming. Although existing instance segmentation networks (such as Maskrcnn) can detect segment fish, most them not effective in monitoring. In order improve image promote accurate intelligent industry, this article uses YOLOv5 backbone network object detection branch, combined with semantic head for segmentation. experiments show that precision reach 95.4% 98.5% algorithm structure proposed article, based on golden crucian carp dataset, 116.6 FPS be achieved RTX3060. On publicly available dataset PASCAL VOC 2007, is 73.8%, 84.3%, speed up 120

Language: Английский

Citations

19
Capsule Network with Its Limitation, Modification, and Applications—A Survey DOI Creative Commons
Mahmood Ul Haq, Muhammad Athar Javed Sethi, Atiq Ur Rehman

et al.

Machine Learning and Knowledge Extraction, Journal Year: 2023, Volume and Issue: 5(3), P. 891 - 921

Published: Aug. 2, 2023

Numerous advancements in various fields, including pattern recognition and image classification, have been made thanks to modern computer vision machine learning methods. The capsule network is one of the advanced algorithms that encodes features based on their hierarchical relationships. Basically, a type neural performs inverse graphics represent object different parts view existing relationship between these parts, unlike CNNs, which lose most evidence related spatial location requires lots training data. So, we present comparative review architectures used applications. paper’s main contribution it summarizes explains significant current published with advantages, limitations, modifications,

Language: Английский

Citations

17
Multi-Level Security in Healthcare by Integrating Lattice-Based Access Control and Blockchain- Based Smart Contracts System DOI Creative Commons
T. Haritha,

A. Anitha

IEEE Access, Journal Year: 2023, Volume and Issue: 11, P. 114322 - 114340

Published: Jan. 1, 2023

Access control to patient information has become increasingly important in healthcare systems. It is vital enhance the security of systems avoid data loss despite various policies imposed by management. The issue needs be resolved with a comprehensive secure framework, which allows users access according their level confidentiality. This article presents solution imposing multi-level e-health integrating Lattice-Based Control (LBAC) model and blockchain-based smart contract mechanisms. These mechanisms provide levels compliance restrictions among resources while maintaining levels. By using LBAC, you can multilevel protection for restrictions, whereas contracts are used ensure transaction process decentralized system via an agreement between parties. A validates every user performs authentication envisioned model, uses Ethereum Virtual Machine (EVM). In blockchain network, patient's details accessed stored as immutable blocks. Comparing proposed scheme existing benchmarking methods reveals that preserves privacy, maintains transparency, provides process, integrity, security. better than other models. As result, lattice-based enhances records.

Language: Английский

Citations

17
Robust Self‐powered Optoelectronic Synapses Based on Epitaxial InSe/GaN Heterojunction with Interfacial Charge‐Trapping Layer DOI

Xuecen Miao,

Yinuo Zhang,

Yunan Lin

et al.

Advanced Optical Materials, Journal Year: 2024, Volume and Issue: 12(19)

Published: April 27, 2024

Abstract Neuromorphic devices that parallelize perception, preprocessing, and computation functions are expected to play a significant role in future non‐von Neumann architecture computers. Herein, new retina‐inspired broadband self‐powered optoelectronic synaptic device based on 2D/3D heterojunction of epitaxial InSe GaN(0001) is reported. Few‐layer n‐type grown p‐type GaN by physical vapor deposition an ultra‐high vacuum (UHV) environment. The fabricated using shadow mask assisted UHV electrode technique. High‐resolution transmission electron microscopy images reveal atomically thin amorphous layer, which induces highly efficient charge trapping, formed at the InSe/GaN interface. photoresponse spans from visible near‐infrared, response time prolonged 10 3 ms owing deep trapping levels. Thus, functions, including excitatory postsynaptic current, paired‐pulse facilitation with high index up 170%, short‐term plasticity, high‐pass filtering characteristics, realized. Additionally, synapses demonstrated merit realizing image sharpening arithmetic operations same under infrared light illumination. This study provides platform heterostructures for robust may find applications post‐Moore era neuromorphic vision systems.

Language: Английский

Citations

5
Multitask learning for image translation and salient object detection from multimodal remote sensing images DOI
Yuanfeng Lian, Xu Shi,

ShaoChen Shen

et al.

The Visual Computer, Journal Year: 2023, Volume and Issue: 40(3), P. 1395 - 1414

Published: May 4, 2023

Language: Английский

Citations

11
A Comprehensive Analysis of Blockchain Applications for Securing Computer Vision Systems DOI Creative Commons

M. Ramalingam,

G. Chemmalar Selvi,

Nancy Victor

et al.

IEEE Access, Journal Year: 2023, Volume and Issue: 11, P. 107309 - 107330

Published: Jan. 1, 2023

Blockchain (BC) and Computer Vision (CV) are the two emerging fields with potential to transform various sectors. BC can offer decentralized secure data storage, while CV allows machines learn understand visual data. The integration of technologies holds massive promise for developing innovative applications that provide solutions challenges in sectors such as supply chain management, healthcare, smart cities, defense. This review explores a comprehensive analysis by examining their combination applications. It also provides detailed fundamental concepts both technologies, highlighting strengths limitations. paper current research efforts make use benefits offered this combination. be used an added layer security systems ensure integrity, enabling image video analytics. open issues associated identified, appropriate future directions proposed.

Language: Английский

Citations

11
MobileOne-YOLO: Improving the YOLOv7 network for the detection of unfertilized duck eggs and early duck embryo development - a novel approach DOI
Qingxu Li,

Ziyan Shao,

Wanhuai Zhou

et al.

Computers and Electronics in Agriculture, Journal Year: 2023, Volume and Issue: 214, P. 108316 - 108316

Published: Oct. 20, 2023

Language: Английский

Citations

11
Вибір ефективної моделі для розпізнавання військових об'єктів у режимі реального часу на спеціалізованих наборах даних DOI Creative Commons

С. І. Глод,

Anastasiya Doroshenko

Scientific Bulletin of UNFU, Journal Year: 2025, Volume and Issue: 35(1), P. 137 - 148

Published: March 6, 2025

Розпізнавання об'єктів у режимі реального часу є ключовим елементом сучасного комп'ютерного зору, особливо в складних сценаріях їх отримання, таких як військові операції, де швидкість і точність виявлення цільових критично важливими для успішної навігації динамічних непередбачуваних умовах поля бою. У цьому дослідженні проаналізовано проблему та класифікації військових часу. Навчено налаштовано три моделі об'єктів: Faster R-CNN (англ. Region-based Convolutional Neural Networks), SSD Single Shot MultiBox Detector) YOLO You Look Only Once). Досліджено продуктивність двоетапних одноетапних алгоритмів й оцінено придатність моделей оперативного розгортання середовищах. Розроблено спеціалізований набір даних, що містить різноманітні зображення бронетехніки (танків, бойових машин піхоти бронетранспортерів) адаптований навчання, валідації тестування реальних умовах. Оцінено навчених за ключовими показниками: точність, влучність, F1-міра, середня частота кадрів. Застосовано платформу NVIDIA Jetson продуктивності умов обмежених обчислювальних ресурсів. Встановлено, модель YOLOv8n найефективнішою, досягнувши найвищих значень mAP (91,8 %) FPS (55), підтверджує її вирішення завдань розпізнавання зображень Водночас, разом із залишковою нейронною мережею ResNet50 Residual Network) забезпечила належну (mAP – 89,2 %, F1-Score 89,4 %), однак низька оброблення вхідних кадрів (FPS 7) значно обмежує використання оперативних сценаріях. Модель з легкою згортковою MobileNetV3 продемонструвала збалансовані результати 81 83,4 36), пропонуючи компроміс між точністю швидкістю, проте поступається загальною ефективністю через випадки хибної або пропуску об'єктів. Вказано на практичну значущість вибору адаптації відповідно до конкретних потреб, зокрема військовій сфері. Отримані слугують основою подальших досліджень, спрямованих вдосконалення часу, розширення набору удосконалення сучасних методів підвищення периферійних пристроїв

Citations

0