| 000 | 04653nas a2200457 a 4500 | ||
|---|---|---|---|
| 005 | 20260818090358.0 | ||
| 008 | 250129s20239999th u ms t 000 eng d | ||
| 035 | _a.b12445952 | ||
| 099 | 9 | _aAIT Diss. no.RS-23-01 | |
| 100 | 1 | _aMaharjan, Nisha | |
| 245 | 1 | 0 | _aDetection of river plastic using UAV sensor data and deep learning |
| 260 |
_aPathum Thani, Thailand : _bAsian Institute of Technology, _c2023 |
||
| 300 |
_a144 leaves : _bill. (some col.)+ _e1 online resource |
||
| 490 | 1 |
_aDissertation ; _vno. RS-23-01 |
|
| 500 | _aA dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Engineering in Remote Sensing and Geographic Information Systems | ||
| 502 | _aThesis (Ph.D.) - Asian Institute of Technology, 2023 | ||
| 520 | _aPlastic pollution has been a critical issue in today's world. It takes up to 1000 years to degrade plastic affecting the natural ecosystem. The river is considered a plastic highway to the ocean and is predicted that there will be more plastic than fish in the ocean by 2050. The monitoring and management of plastic in rivers has been essential. However, in-situ plastic observation by humans is tedious and cumbersome. This research presents plastic mapping in rivers integrating Unmanned Aerial Vehicle and deep learning (DL) in the Houay Mak Hiao (HMH) river of Laos together with Khlong Nueng canal of Talad Thai (TT), Thailand. Pre-trained neural networks ResNet-101 and GoogLeNet are used to classify plastic and non-plastic in the images with classification accuracy for HMH is 83.6% and 82.9% while for TT is 86.1% and 87.4% respectively. The respective validation accuracy using ResNet-101 and GoogLeNet for HMH are 86% and 85.33% and for TT is 89.33% and 90% respectively. The precision of detection at the Intersection over Union (ToU) threshold 0.5 using You Only Look Once (YOLO)v2 with ResNet-101 and GoogLeNet for HMH are 0.76 and 0.67 and for TT is 0.75 and 0.72 respectively. The transferability of the trained model from HMH to TT and vice-versa is assessed with kappa co-efficient 0.25 in classification task from HMH to TT while only 0.12 in the detection, showing classification task wins. GoogLeNet has exceptionally high kappa co-efficient 0.42 in case of model transfer from TT to HMH which may be chance, but classification put the image conditionally competitive than detection saving computing resources and time for data preparation and training. This research presents plastics detection methods on ortho imageries using pre-trained YOLO family, scratch, and their transferability. Pre-trained YOLOV5s show higher performance than other models with mean Average Precision (mAP) of 0.81 at IoU 0.5 without transfer learning. Pre-trained YOLOv4 with transfer learning, however, demonstrates the highest accuracy, with a 3.0% boost in mAP reach 0.83, compared to a negligible increase of roughly 2% for pre-trained YOLOv5s. Following transfer learning from TT to HMH, there was increase in mAP from 0.59 to 0.81. YOLOV3 from scratch exhibits the greatest boost via transfer learning. The various stakcholders get the benefit of knowledge building regarding monitoring of plastic, provisions,and challenges in understanding the latest technology for plastic monitoring. | ||
| 650 | 0 |
_aPlastics _xEnvironmental aspects _xData processing |
|
| 650 | 0 | _aEnvironmental monitoring | |
| 650 | 0 | _aDeep learning (Machine learning) | |
| 650 | 0 | _aPlastic pollution | |
| 700 | 1 |
_aMiyazaki, Hiroyuki, _eChairperson |
|
| 700 | 1 |
_aShrestha, Sangam, _eCo-chairperson |
|
| 700 | 1 |
_aNakamura, Tai, _eExamination Committee |
|
| 700 | 1 |
_aDailey, Matthew N., _eExamination Committee |
|
| 710 | 2 |
_aGovernment of Japan, _eScholarship Donor |
|
| 810 | 2 |
_aAsian Institute of Technology. _tDissertation ; _vno. RS-23-01 |
|
| 856 |
_3Full-Text _uhttp://203.159.5.9/ait-thesis/detail.php?q=B23586 |
||
| 907 |
_a.b12445952 _bmnarc _ca |
||
| 902 | _a260225 | ||
| 998 |
_b0 _c260223 _dm _ea _fa _g0 |
||
| 945 | _lmnarc | ||
| 945 | _lmnait | ||
| 945 | _lmnait | ||
| 942 | _c40 | ||
| 942 | _c20 | ||
| 909 |
_aBarcode : 30050120902043 _bCREATED : 2025-01-29 _cRECORD # : i13530616 _dLPATRON : 0 _eLCHKIN : - _f# RENEWALS : 0 _g# OVERDUE : 0 _hIUSE3 : 0 _iTOT CHKOUT : 0 _jTOT RENEW : 0 |
||
| 909 |
_aBarcode : 30050120424253 _bCREATED : 2025-09-09 _cRECORD # : i13556290 _dLPATRON : 0 _eLCHKIN : - _f# RENEWALS : 0 _g# OVERDUE : 0 _hIUSE3 : 0 _iTOT CHKOUT : 0 _jTOT RENEW : 0 |
||
| 909 |
_aBarcode : 30050120424246 _bCREATED : 2025-09-09 _cRECORD # : i13556307 _dLPATRON : 0 _eLCHKIN : - _f# RENEWALS : 0 _g# OVERDUE : 0 _hIUSE3 : 0 _iTOT CHKOUT : 0 _jTOT RENEW : 0 |
||
| 999 |
_c27895 _d27895 |
||