黑料不打烊

I拧maniosios lempos nuovargio aptikimui k奴rimas. Bakalauro baigiamasis darbas elektronikos in啪inerijos laipsniui. Vilniaus Gedimino technikos universitetas. Vilnius, 2026, 55p., 10 iliustr., 2 lent., 21 bibl., 7 pried懦. 艩iame bakalauro baigiamajame darbe nagrin臈jamas i拧maniosios lempos nuovargio aptikimui sistemos k奴rimas. Darbo tikslas 鈥� sukurti realiuoju laiku veikian膷i膮 sistem膮, geban膷i膮 steb臈ti vartotojo aki懦 b奴sen膮 ir persp臈ti apie aptiktus nuovargio po啪ymius. Sistema paremta kompiuterin臈s regos metodais, naudojant mini kompiuter寞 鈥濺aspberry Pi鈥� ir ESP32-CAM kamer膮. Darbe sukurti aki懦 mirks臈jimo aptikimo ir nuovargio nustatymo algoritmai, paremti aki懦 atv臈rimo santykio analize, mirks臈jimo da啪nio bei aki懦 u啪merkimo trukm臈s vertinimu. Siekiant prisitaikyti prie skirting懦 vartotoj懦 ir darbo s膮lyg懦, 寞gyvendinti rankinis ir adaptyvus veikimo re啪imai. Taip pat realizuota fizin臈 sistema su servo mechanizmu kameros sekimui, rel臈s valdymu bei vartotojo grafine s膮saja. Eksperimentinio testavimo metu buvo vertinamas sistemos geb臈jimas aptikti aki懦 mirks臈jimus ir reaguoti 寞 nuovargio po啪ymius realiomis s膮lygomis. Gauti rezultatai parod臈, kad sistema daugeliu atvej懦 patikimai aptinka aki懦 b奴senos poky膷ius, o vidutinis mirks臈jimo aptikimo tikslumas siek臈 apie 90 %. Darbo rezultatai patvirtino, kad sukurta sistema atitinka i拧kelt膮 tiksl膮 ir gali b奴ti pritaikoma praktiniam naudojimui vartotojo budrumo steb臈senai.

Development of a Smart Lamp for Fatigue Detection. Bachelor鈥檚 Final Thesis for the Degree of Electronics Engineering. 黑料不打烊_独家黑料_吃瓜网51爆料. Vilnius, 2026, 55 pages, 10 illustrations, 2 tables, 21 references, 7 appendices. This Bachelor鈥檚 thesis analyzes the development of a smart lamp system for fatigue detection. The aim of the thesis is to develop a real-time system capable of monitoring the user鈥檚 eye state and providing warnings when signs of fatigue are detected. The system is based on computer vision methods using a Raspberry Pi mini computer and an ESP32-CAM camera. Blink detection and fatigue identification algorithms were developed based on eye aspect ratio analysis, blink frequency, and eye closure duration evaluation. In order to adapt to different users and working conditions, both manual and adaptive operating modes were implemented. A physical system with a servo mechanism for camera tracking, relay control, and a graphical user interface was also realized. During experimental testing, the system鈥檚 ability to detect eye blinks and respond to signs of fatigue under real conditions was evaluated. The obtained results showed that the system reliably detects changes in eye state in most cases, while the average blink detection accuracy reached approximately 90%. The results of the thesis confirmed that the developed system meets the defined objective and can be applied for practical user alertness monitoring.

Vie拧ojo saugumo steb臈jimo sistemose augant kamer懦 skai膷iui ir vaizdo duomen懦 srautams, rankin臈 analiz臈 tampa nepakankama: vienas operatorius efektyviai stebi tik apie 10鈥�20 kamer懦. Darbo tikslas 鈥� sukurti ir i拧tirti automatizuot膮 啪moni懦 grupi懦 elgsenos atpa啪inimo i拧 vaizdo algoritm膮. 漠gyvendinta keturi懦 etap懦 apdorojimo grandin臈: asmen懦 aptikimas Ultralytics YOLO detektoriumi, keli懦 objekt懦 sekimas BoT-SORT algoritmu, grupi懦 formavimas DBSCAN metodu su dinaminiu spinduliu ir elgsenos klasifikavimas atskiru YOLO-cls modeliu, pasitelkiant 10 klasi懦 taksonomij膮. Baziniame eksperimento variante taikytas yolo26x nustatymas (conf=0,22, imgsz=960), o papildomuose bandymuose vertinti ir alternatyv奴s nustatymai. Geriausias pasiektas klasifikatoriaus tikslumas validacijos rinkinyje 鈥� 0,7243 (top-1). Aptikimo etape mAP@0,5 siek臈 0,719. Sekimo kokyb臈, 寞vertinta su anotuotu 寞ra拧u, buvo itin auk拧ta: MOTA = 0,9974, IDF1 = 0,9997. Sistema apdoroja vaizdo sraut膮 25,8鈥�26,1 kadr懦 per sekund臋 grei膷iu, atitinkan膷iu realiojo laiko reikalavimus. Tankiose miniose aptikimo jautrumas suma啪臈ja, o grupi懦 formavimas tampa nestabilus 鈥� tai pagrindin臈 identifikuota sistemos riba. Darb膮 sudaro 8 dalys: 寞vadas, analitin臈 literat奴ros ap啪valga, 啪moni懦 grup臈s elgsenos vaizde atpa啪inimo algoritmo 寞gyvendinimas, duomen懦 rinkimas ir parengimas eksperimentams, 啪moni懦 grupi懦 elgsenos atpa啪inimo algoritmo k奴rimas, 啪moni懦 grupi懦 elgsenos vaizde atpa啪inimo tyrimo rezultatai, i拧vados, literat奴ros s膮ra拧as. Darbo apimtis 鈥� 63 p. teksto be pried懦, 8 iliustr., 11 lent., 37 bibliografiniai 拧altiniai. Atskirai pridedami darbo priedai.

As the number of surveillance cameras and video streams increases in public-safety systems, manual monitoring becomes insufficient: a single operator can effectively supervise only about 10鈥�20 cameras. The aim of this thesis is to develop and investigate an automated algorithm for group human-behavior recognition from video. A four-stage pipeline was implemented: person detection with an Ultralytics YOLO detector, multi-object tracking with BoT-SORT, group formation via DBSCAN with a dynamic grouping radius, and group-behavior classification with a separate YOLO-cls model using a 10-class taxonomy. In the baseline setup, the yolo26x preset (conf=0.22, imgsz=960) was used, while alternative presets were evaluated in additional runs. The best achieved classifier accuracy on the validation set was 0.7243 (top-1), and detection mAP@0.5 reached 0.719. Tracking quality on an annotated recording was very high: MOTA = 0.9974 and IDF1 = 0.9997. The system processes video at 25.8鈥�26.1 FPS, meeting real-time requirements. In dense crowds, detection recall drops and group formation becomes unstable 鈥� this is the primary identified limitation of the system. The thesis consists of 8 parts: introduction, analytical literature review, implementation of the algorithm for recognising the behaviour of a group of people in video, data collection and preparation for experiments, development of the algorithm for recognising the behaviour of groups of people, results of the study on recognising the behaviour of groups of people in video, conclusions, and list of references. The thesis comprises 63 pages of text without appendices, 8 illustrations, 11 tables, and 37 bibliographic sources. Thesis appendices are attached separately.

I拧tirti dirbtinio intelekto modeli懦 taikymo galimybes triuk拧mo ma啪inimui elektronin臈se grandin臈se ir 寞vertinti j懦 efektyvum膮 signal懦 apdorojime. Tikslui pasiekti buvo atlikta mokslin臈s literat奴ros analiz臈, i拧nagrin臈ti triuk拧mo 拧altiniai bei tradiciniai ir dirbtiniu intelektu pagr寞sti triuk拧mo slopinimo metodai. Darbe pasirinkti ir realizuoti ma拧ininio mokymosi modeliai, pritaikyti reali懦 signal懦 filtravimui. Sukurtas eksperimentinis sprendimas buvo testuojamas su triuk拧mu u啪ter拧tais signalais, vertinant triuk拧mo suma啪inimo lyg寞 ir signal懦 kokyb臈s pager臈jim膮. Gauti rezultatai parod臈, kad taikyti dirbtinio intelekto modeliai efektyviai suma啪ina triuk拧m膮 ir pagerina signal懦 charakteristikas. Tai patvirtina, kad darbo tikslai pasiekti, o sukurti sprendimai s臈kmingai tenkina i拧kelt膮 u啪duot寞 Darb膮 sudaro 1. 漠vadas; 2. Dirbtiniu intelektu pagr寞stu triuk拧mo slopinimo metod懦 analitin臈 literat奴ros ap啪valga; 3.Triuk拧mo slopinimo metod懦 analiz臈; 4. Triuk拧mo slopinimo duomen懦 atranka; 5.Eksperimentinio tyrimo rezultat懦 analiz臈; 6.Apibendrinimas. I拧vados; Literat奴ra; Priedai Darbo apimtis - 76 p. teksto be pried懦, 13 iliustracij懦, 6 lentel臈s, 21 bibliografiniai 拧altiniai, 10 pried懦- 79 p

The possibilities of applying artificial intelligence models for noise reduction in electronic circuits were investigated and their effectiveness in signal processing was evaluated. To achieve this goal, an analysis of scientific literature was conducted, noise sources and both traditional and artificial intelligence-based noise suppression methods were examined. Machine learning models were selected and implemented, adapted for the filtering of real signals. The developed experimental solution was tested with noise-contaminated signals, evaluating the level of noise reduction and the improvement of signal quality. The results obtained demonstrated that the applied artificial intelligence models effectively reduce noise and improve signal characteristics. This confirms that the objectives of the work have been achieved, and the developed solutions successfully fulfil the stated task. The work consists 9 parts: 1. Introduction; 2. Analytical literature review of artificial intelligence-based noise suppression methods; 3. Analysis of noise suppression methods; 4. Selection of noise suppression data; 5. Analysis of experimental research results; 6. Summary. Conclusions; References; Appendices. Thesis consist of: 76 pages of text excluding appendices, 13 illustrations, 6 tables, 21 bibliographic sources, 10 appendices 鈥� 79 pages.

ARM architekt奴ros mikroklasteriui specializuota atvirojo kodo sistema. Bakalauro baigiamasis darbas informatikos in啪inerijos laipsniui. Vilniaus Gedimino technikos universitetas. Vilnius, 2026, 114 p., 55 iliustr., 15 lent., 39 bibl., 1 pried懦. Bakalauro baigiamojo darbo tikslas 鈥� suprojektuoti ir 寞gyvendinti atvirojo kodo sistem膮, leid啪ian膷i膮 valdyti virtualizuotas operacines sistemas trij懦 mazg懦 ARM architekt奴ros mikroklasteryje. Atlikta analogi拧k懦 virtualizacijos ir konteinerizacijos sprendim懦 ap啪valga, palygintos pramon臈s standartu laikomos platformos su lengvasvor臈mis alternatyvomis bei i拧skirti j懦 privalumai ir tr奴kumai ARM platformos kontekste. Apibr臈啪ti sistemos reikalavimai techninei 寞rangai, vietiniam tinklui ir programinei daliai. Remiantis aptartais reikalavimais parinkta 鈥瀔3s" konteinerizacijos platforma, integruota su 鈥濳ubeVirt" virtualizacijos 寞skiepiu, 鈥濴onghorn" paskirstyt膮ja saugykla bei 鈥濸rometheus" ir 鈥濭rafana" steb臈senos sistema. Visas diegimo procesas automatizuotas pagal infrastrukt奴ros kaip kodo principus naudojant 鈥濧nsible" scenarijus. I拧tirtos sukurtos sistemos atsparumo trikd啪iams ir na拧umo charakteristikos, palygintos su u啪duoties s膮lygos reikalavimais. Apibendrinti darbo rezultatai i拧vadose

ARM Architecture Microcluster based Open-Source System. Bachelor thesis in informatics engineering. Vilnius Tech University. Vilnius, 2026, 114 p., 55 illustrations, 15 tables, 39 bibliographical sources, 1 attachment. The aim of the bachelor thesis is to design and implement an open-source system for managing virtualized operating systems on a three-node ARM architecture microcluster. The work includes a review of analogous virtualization and containerization solutions, comparing industry-standard platforms with lightweight alternatives and identifying their advantages and disadvantages in the context of the ARM platform. Requirements for hardware, local area network and software components were defined. Based on those requirements, the k3s container orchestration platform was selected, integrated with the KubeVirt virtualization extension, Longhorn distributed storage, and a Prometheus and Grafana monitoring stack. The entire deployment process is automated according to infrastructure-as-code principles using Ansible playbooks. The fault tolerance and performance characteristics of the developed system were investigated and compared against the task requirements. The final results are summarized in the conclusion.

I拧mani ant啪emini懦 jutikli懦 tinklo steb臈senos ir nuotolinio valdymo sistema. Bakalauro baigiamasis darbas informatikos in啪inerijos laipsniui. Vilnius Tech universitetas. Vilnius, 2026, 63 p., 14 iliustr., 11 lent., 12 bibl., 7 priedai. Darbo tikslas 鈥� sukurti i拧mani膮 ant啪emini懦 jutikli懦 tinklo steb臈senos ir nuotolinio valdymo sistem膮. Tikslui pasiekti suprojektuota daugiasluoksn臈 architekt奴ra, duomen懦 baz臈 ir nar拧yklin臈 operatoriaus s膮saja. 漠gyvendinta serverin臈 dalis, realiojo laiko ry拧ys, jutikli懦 duomen懦 apdorojimas ir 啪em臈lapio vaizdavimas. Sistema priima 鈥濻ensor Server鈥� prane拧imus, formuoja individualius bei grupinius incidentus ir leid啪ia nuotoliniu b奴du keisti jutikli懦 parametrus. Testavimas parod臈, kad rezultatai atitinka u啪duotyje numatytus steb臈senos, analiz臈s ir valdymo reikalavimus.

Intelligent Unattended Ground Sensor Network Monitoring and Remote Management System. Bachelor鈥檚 thesis for the degree in Informatics Engineering. Vilnius Tech University. Vilnius, 2026, 63 pages, 14 figures, 11 tables, 12 bibliographic references, 7 appendices. The aim of the thesis is to create an intelligent unattended ground sensor network monitoring and remote management system. To achieve this aim, a multilayer architecture, database and browser-based operator interface were designed. The server side, real-time communication, sensor data processing and map visualization were implemented. The system receives Sensor Server messages, forms individual and group incidents, and allows sensor parameters to be changed remotely. Testing showed that the results meet the monitoring, analysis and control requirements defined in the task.

Baigiamajame bakalauro darbe sukurta interaktyvaus v臈jo j臈gain臈s skaitmeninio dvynio sistema gedim懦 prevencijai bei nustatyti tokios sistemos veikimo ypatumai ir taikymo privalumai. Darbo tikslas 鈥� sukurti interaktyv懦 v臈jo j臈gain臈s skaitmenin寞 dvyn寞 su jutikli懦 sistema ir nuotolinio valdymo galimybe, 寞gyvendinant realaus laiko duomen懦 steb臈jim膮 bei mechanini懦 gedim懦 ir strukt奴rini懦 anomalij懦 diagnostik膮. Darbo analitin臈je dalyje buvo nagrin臈jami fizini懦 v臈jo j臈gaini懦 maket懦 analogai, atliktas mikrovaldikli懦 platform懦 (鈥濺aspberry Pi鈥�, 鈥濧rduino Uno鈥�, 鈥濫SP-32鈥�) technini懦 specifikacij懦 palyginimas, inercini懦 bei kampo sekimo jutikli懦 lyginamoji analiz臈 bei parinkti tinkamiausi elektromechaniniai komponentai. Projektavimo ir ekonominio pagrindimo dalyje buvo atlikta SSGG (SWOT) analiz臈, parengta sistemos strukt奴rin臈 ir duomen懦 sraut懦 funkcin臈 schema, suprojektuota laiko eilu膷i懦 duomen懦 baz臈s lenteli懦 strukt奴ra bei modeliuojami keturi kritiniai anomalij懦 aptikimo scenarijai. Darbo praktin臈je dalyje buvo sukonstruotas fizinis 3D maketas i拧 PLA plastiko su integruotais jutikliais (MPU6050, AS5600) bei realizuota programin臈 寞ranga 鈥濫SP-32鈥� mikrovaldiklio lygmenyje. Sistemos programin臈 ekosistema ir valdymo pultas realizuoti naudojant 鈥濶ext.js鈥� karkas膮 bei laiko eilu膷i懦 duomen懦 baz臋 鈥濱nfluxDB Cloud鈥�. Sistemos testavimas buvo atliktas laboratorin臈je aplinkoje testuojant j臈gain臋 statin臈je ramyb臈s b奴senoje, dinaminio azimuto keitimo metu bei modeliuojant kritines mechanines anomalijas ir avarin寞 stabdym膮.

In this bachelor's thesis, an interactive digital twin system for a wind turbine was developed for fault prevention, and the operating characteristics and application benefits of this system were determined. The aim of the thesis is to create an interactive wind turbine digital twin with a sensor system and remote control capability, implementing real-time data monitoring and diagnostics of mechanical faults and structural anomalies. In the analytical part of the thesis, physical wind turbine mock-up analogues were analyzed, a technical specification comparison of microcontroller platforms (Raspberry Pi, Arduino Uno, ESP-32) was performed, a comparative analysis of inertial and yaw-tracking sensors was conducted, and the most suitable electromechanical components were selected. In the design and economic justification part, a SWOT analysis was conducted, the system's structural and data flow functional diagrams were prepared, the table structure of the time-series database was designed, and four critical anomaly detection scenarios were modeled. In the practical part of the thesis, a physical 3D mock-up was constructed from PLA plastic with integrated sensors (MPU6050, AS5600), and the software was implemented at the ESP-32 microcontroller level. The software ecosystem and control dashboard were implemented using the Next.js framework and the InfluxDB Cloud time-series database. System testing was conducted in a laboratory environment, evaluating the turbine in a static idle state, during dynamic yaw changes, and under simulated critical mechanical anomalies and emergency stops.

Baigiamajame bakalauro darbe sukurta interaktyvaus v臈jo j臈gain臈s skaitmeninio dvynio sistema gedim懦 prevencijai bei nustatyti tokios sistemos veikimo ypatumai ir taikymo privalumai. Darbo tikslas 鈥� sukurti interaktyv懦 v臈jo j臈gain臈s skaitmenin寞 dvyn寞 su jutikli懦 sistema ir nuotolinio valdymo galimybe, 寞gyvendinant realaus laiko duomen懦 steb臈jim膮 bei mechanini懦 gedim懦 ir strukt奴rini懦 anomalij懦 diagnostik膮. Darbo analitin臈je dalyje buvo nagrin臈jami fizini懦 v臈jo j臈gaini懦 maket懦 analogai, atliktas mikrovaldikli懦 platform懦 (鈥濺aspberry Pi鈥�, 鈥濧rduino Uno鈥�, 鈥濫SP-32鈥�) technini懦 specifikacij懦 palyginimas, inercini懦 bei kampo sekimo jutikli懦 lyginamoji analiz臈 bei parinkti tinkamiausi elektromechaniniai komponentai. Projektavimo ir ekonominio pagrindimo dalyje buvo atlikta SSGG (SWOT) analiz臈, parengta sistemos strukt奴rin臈 ir duomen懦 sraut懦 funkcin臈 schema, suprojektuota laiko eilu膷i懦 duomen懦 baz臈s lenteli懦 strukt奴ra bei modeliuojami keturi kritiniai anomalij懦 aptikimo scenarijai. Darbo praktin臈je dalyje buvo sukonstruotas fizinis 3D maketas i拧 PLA plastiko su integruotais jutikliais (MPU6050, AS5600) bei realizuota programin臈 寞ranga 鈥濫SP-32鈥� mikrovaldiklio lygmenyje. Sistemos programin臈 ekosistema ir valdymo pultas realizuoti naudojant 鈥濶ext.js鈥� karkas膮 bei laiko eilu膷i懦 duomen懦 baz臋 鈥濱nfluxDB Cloud鈥�. Sistemos testavimas buvo atliktas laboratorin臈je aplinkoje testuojant j臈gain臋 statin臈je ramyb臈s b奴senoje, dinaminio azimuto keitimo metu bei modeliuojant kritines mechanines anomalijas ir avarin寞 stabdym膮. Baigiam膮j寞 darb膮 sudaro: 寞vadas, v臈jo j臈gain臈s maketo analog懦 ir komponent懦 ap啪valga, skaitmeninio dvynio sistemos ekonominis pagrindimas ir projektavimas, skaitmeninio dvynio fizinio maketo ir programin臈s dalies 寞gyvendinimas, sistemos patikra bei rezultat懦 apibendrinimas. Darbo apimtis 鈥� 60 puslapiai be pried懦, 8 lentel臈s bei 22 iliustracijos. Naudojamasi 21 literat奴ros 拧altiniu.

In this bachelor's thesis, an interactive digital twin system for a wind turbine was developed for fault prevention, and the operating characteristics and application benefits of this system were determined. The aim of the thesis is to create an interactive wind turbine digital twin with a sensor system and remote control capability, implementing real-time data monitoring and diagnostics of mechanical faults and structural anomalies. In the analytical part of the thesis, physical wind turbine mock-up analogues were analyzed, a technical specification comparison of microcontroller platforms (Raspberry Pi, Arduino Uno, ESP-32) was performed, a comparative analysis of inertial and yaw-tracking sensors was conducted, and the most suitable electromechanical components were selected. In the design and economic justification part, a SWOT analysis was conducted, the system's structural and data flow functional diagrams were prepared, the table structure of the time-series database was designed, and four critical anomaly detection scenarios were modeled. In the practical part of the thesis, a physical 3D mock-up was constructed from PLA plastic with integrated sensors (MPU6050, AS5600), and the software was implemented at the ESP-32 microcontroller level. The software ecosystem and control dashboard were implemented using the Next.js framework and the InfluxDB Cloud time-series database. System testing was conducted in a laboratory environment, evaluating the turbine in a static idle state, during dynamic yaw changes, and under simulated critical mechanical anomalies and emergency stops. The bachelor's thesis consists of: an introduction, an overview of wind turbine mock-up analogues and components, the economic justification and design of the digital twin system, the implementation of the physical mock-up and software of the digital twin system, system validation, and a summary of results. The scope of the thesis is 60 pages excluding appendices, containing 8 tables and 22 illustrations. It references 21 literature sources.

艩iame bakalauro baigiamajame darbe sukurta automobili懦 stov臈jimo aik拧tel臈s u啪imtumo analiz臈s sistema, skirta realiuoju laiku aptikti automobilius ir 啪mones bei analizuoti aik拧tel臈s apkrov膮. Sistema paremta kompiuterinio reg臈jimo ir giluminio mokymosi technologijomis. Objekt懦 aptikimui naudojami YOLO modeliai, o objekt懦 sekimui naudojamas 鈥濨yteTrack鈥� algoritmas. Vaizdo apdorojimas vykdomas naudojant 鈥濶VIDIA Jetson Nano鈥� platform膮. Sukurta sistema analizuoja vaizdo sraut膮 i拧 RGB arba termin臈s kameros, aptinka objektus, apskai膷iuoja j懦 kiek寞 ir perduoda duomenis 寞 鈥濬irebase Firestore鈥� duomen懦 baz臋. Duomen懦 perdavimui ir vaizdo transliacijai naudojamas 鈥濬lask鈥� serveris. Taip pat sukurta WEB vartotojo s膮saja, leid啪ianti steb臈ti automobili懦 ir 啪moni懦 skai膷i懦, analizuoti istorinius duomenis bei realiuoju laiku matyti apdorot膮 vaizdo sraut膮. Darbe atlikta esam懦 automobili懦 stov寞jimo aik拧teli懦 steb臈jimo sistem懦 analiz臈, suprojektuota sistemos architekt奴ra, realizuoti pagrindiniai programiniai moduliai ir atliktas sistemos veikimo testavimas. Gauti rezultatai parod臈, kad sistema veikia stabiliai, u啪tikrina realaus laiko objekt懦 aptikim膮 bei leid啪ia efektyviai analizuoti aik拧tel臈s u啪imtum膮.

This bachelor thesis presents the development of a parking occupancy analysis system designed for real-time vehicle and pedestrian detection and parking load analysis. The system is based on computer vision and deep learning technologies. Yolo models are used for object detection, while the ByteTrack algorithm is applied for object tracking. Video processing is performed using the NVIDIA Jetson Nano platform. The developed system analyzes video streams from RGB or thermal cameras, detects objects, calculates their quantity, and transfers the collected data to the Firebase Firestore database. A Flask server is used for data transmission and live video streaming. In addition, a WEB user interface was developed to display the number of detected vehicles and people, visualize historical data, and provide real-time processed video streaming. The thesis includes an analysis of existing parking monitoring systems, system architecture design, implementation of the main software modules, and system testing. The obtained results demonstrated that the system operates reliably, ensures real-time object detection, and enables efficient parking occupancy analysis.

Bakalauro baigiamojo darbo metu projektuojama ir realizuojama mobili i拧manioji sveikatos steb臈senos sistema. Sudaryt膮 sistem膮 sudaro i拧manioji apyrank臈, kuri skirta vartotoj懦 sveikatos parametr懦 fiksavimui. Sistemos sud臈tyje yra paciento mobilusis i拧manusis telefonas, skirtas duomen懦 apdorojimui ir slaugytojo telefonas, kurio paskirtis yra steb臈ti pacient懦 parametrus ir gauti prane拧imus esant reik拧mingiems paciento sveikatos parametr懦 pakitimams. Taip pat sistem膮 sudaro Firestore duomen懦 baz臈, kuri saugo nuolatinius ir kintamus vartotoj懦 parametrus bei u啪tikrina vartotoj懦 autentifikavim膮. Darbe sudarytos sistemos architekt奴ros diagramos, apib奴dinan膷ios sistemos sud臈t寞, UML diagramos, kurios apib奴dinim膮 sistemos vykdom懦 proces懦 eig膮. Baigiamojo darbo metu i拧analizuotos analogi拧kos sistemos ir pasirinkti konkret奴s komponentai, kurie naudojami i拧maniosios apyrank臈s projektavimui ir 寞gyvendinimui. Naudojant Angular ir Capacitor bibliotekas sukuriama i拧manioji program臈l臈 skirta paciento steb臈senai. Realizavus bendr膮 sistem膮 atlikti testavimo scenarijai, kuri懦 metu pasteb臈ta, kad gamybos metu, spausdintin臈 plok拧t臈je sudarytas trumpasis jungimas, tod臈l sveikatos steb臈senos sistema realizuota monta啪ine plok拧te ir analogi拧kais jutikliais. Programin臈s 寞rangos testavimo metu pasteb臈ti tr奴kumai, kuriuos panaikinus galima patobulinti vartotojo s膮sajos patogum膮. 漠vertinus testavimo rezultatus, padaryta i拧vada, kad suprojektuota sistema neatitinka konstrukcini懦 parametr懦, ta膷iau sistemos funkcionalumas veikia tinkamai ir neturi trukd啪i懦, kurie trukdyt懦 sistemos veikimui. Mobilioji i拧mani sveikatos steb臈senos sistema.Bakalauro baigiamasis darbas elektronikos in啪inerijos laipsniui. Vilniaus Gedimino technikos universitetas. Vilnius, 2026, 113 p., 50 iliustr., 13 lent., 7 priedai, 3xA4 br臈啪., 1xA3 plak.

Bachelor's thesis involves the design and implementation of a mobile smart health monitoring system. The system consists of a smart wristband designed to track users' health parameters. The system includes the patient鈥檚 mobile smartphone, used for data processing, and the nurse鈥檚 phone, which is used to monitor patient parameters and receive notifications in the event of significant changes in the patient鈥檚 health parameters. The system also includes a Firestore database, which stores static and dynamic user parameters and ensures user authentication. The thesis includes system architecture diagrams describing the system鈥檚 composition, as well as UML diagrams illustrating the flow of processes executed by the system. During the thesis, similar systems were analyzed, and specific components were selected for the design and implementation of the smart bracelet. Using Ionic development tools, a smart app for patient monitoring was created. After implementing the overall system, testing scenarios were performed, during which it was observed that a short circuit had formed on the printed circuit board during manufacturing; therefore, the health monitoring system was implemented using a prototype board and analogous sensors. Deficiencies were identified during software testing; addressing these can improve the user interface鈥檚 usability. After evaluating the test results, it was concluded that the designed system does not meet the design parameters, but the system is functioning properly and has no issues that would interfere with the system's operation. Intelligent mobile health monitoring system. Bachelor thesis for electronics engineering degree. Vilnius Gediminas technical university. Vilnius, 2026, 113 p., 50 illust., 13 tab., 7 attachments, 3xA4 sch., 1xA3 post.

艩iame baigiamajame darbe projektuojama ir realizuojama vaizdo ir ta拧k懦 debesies sutapdinimo sistema, skirta vaizdo kameros pad臈ties ir orientacijos nustatymui trimat臈je aplinkoje pagal vien膮 RGB vaizd膮 ir i拧 anksto rekonstruot膮 3D model寞. Darbe analizuojami vizualin臈s lokalizacijos metodai, taip pat po啪ymi懦 i拧skyrimo algoritmai bei 2D鈥�3D atitikmen懦 paie拧kos principai. Sistemos realizacijai naudota COLMAP sistema, leid啪ianti atlikti po啪ymi懦 i拧skyrim膮, vaizd懦 tarpusavio sutapdinim膮 ir aplinkos 3D rekonstrukcij膮. Taip pat realizuota lokalizacijos rezultat懦 vizualizacija trimat臈je erdv臈je bei vartotojo s膮saja testavimui. Eksperiment懦 metu nustatyta, kad sistema gali s臈kmingai automati拧kai lokalizuoti vaizdo kamer膮 rekonstruotoje aplinkoje, kai pateikiamos pakankamos kokyb臈s ir aplink膮 atitinkan膷ios nuotraukos.

This final thesis presents the design and implementation of an image-to-point cloud alignment system intended for camera pose estimation in a 3D environment using a single RGB image and a previously reconstructed 3D model. The thesis analyzes visual localization methods, feature extraction algorithms, and principles of 2D鈥�3D feature matching. The COLMAP system was used for feature extraction, image matching, and 3D reconstruction of the environment. The system also includes 3D visualization of localization results and a user interface for testing. Experimental results demonstrated that the system is capable of successfully automatically localizing the camera in the reconstructed environment when sufficient image quality and matching environmental features are provided.