NATIONAL UNIVERSITY HO CHI MINH CITY UNIVERSITY OF INFORMATION TECHNOLOGY ADVANCED PROGRAM IN INFORMATION SYSTEMS NGUYEN VAN TUAN - 18521606 CLASSIFICATION OF SATELLITE IMAGES FOR VIETNAM'S INFRASTRUCTURE BACHELOR OF ENGINEERING IN INFORMATION SYSTEMS THESIS ADVISOR PhD. DO TRONG HOP PhD. TRAN VAN THANH HO CHI MINH CITY, 2024 ASSESSMENT COMMITTEE The Assessment Committee is established under the Decision. by Rector of the University of Information Technology.
THESIS ADVISOR'S COMMENTS 900000000 000060606060600000000000000000000000000000000000000000000000000000000000000000000060606s2°°°°°°6°6 90000000 00000000000060000000000000000000000000000000000000000000000000000000000000000060606060 E DEED EEOES 9 0000000000000000000000000000000000000000000000000000000000000000000000000000000000006060606022°°°°°°6°6 90000000000 000000000000000000000000000000000000000000000000000000000000000000006006060606060606066066060662°°°° 66 9 0000000000000000000000000000000000000000000000000909009006060600006000000000000000000606060606060606060060e°°°66 9 0090000000000000000000000000000000000000000000000009009060606000000000000000000000006060606060606066060600606e666 9000000000000 0000000000000000000000000/00000000000000000000000000060000006000000000000606060606060 0606262 errr) 90/00 0000000000000000000000000000000000000000000000000090090000000000000000000000000000060606066006062e°°°e°6°6 9 0000000000000000000000000000000000000000000000000009090909060060060000000000000000000606060606060606606606600°°°°6°6 900000000000000000000009000000000000000600000000000009090090000000000000000000000600000060606060°°eeeeee°66 ĐÔ ÔаĐÔ°ÔeÔ°°eÔC°GCÔC°°bÔeÔeÔbCÔe°o°od°0ôe0ôeôo°od°oô°eôoôeoôo°oô°oôôeoôoôoôeoeoo°o°o°oôoô°eeoeoôoôôeeoo°oôoôoô°eoeeo°o°oôo°eoeo°oôoôe°edeeee°o°oôeô°eeo°oôeeeee°eoeeeoeee°ee°e©°e© *0000000000000000000000000000000000006000060606060000000000000606060606060060600606000000000000000606060606066066°°°°e°6°e6 9 0/00000000000000000000000000000000000000000000000000000000060006060060600000000000000000006060606600606°6°°°°C°e°e6 Ho Chi Minh city,. THESIS ADVISOR THESIS REVIEWER 'S COMMENTS 90000000000 000000000000000000000000000000000000000000000060606000060000000000000000006060606060606066066060660°°°°6°6 900000000 000060606060600000000000000000000000000000000000000000000000000000000000000000000060606s2°°°°°°6°6 90000000 00000000000060000000000000000000000000000000000000000000000000000000000000000060606060 E DEED EEOES 9 0000000000000000000000000000000000000000000000000000000000000000000000000000000000006060606022°°°°°°6°6 90000000000 000000000000000000000000000000000000000000000000000000000000000000006006060606060606066066060662°°°° 66 9 0000000000000000000000000000000000000000000000000909009006060600006000000000000000000606060606060606060060e°°°66 9 0090000000000000000000000000000000000000000000000009009060606000000000000000000000006060606060606066060600606e666 9000000000000 0000000000000000000000000/00000000000000000000000000060000006000000000000606060606060 0606262 errr) 90/00 0000000000000000000000000000000000000000000000000090090000000000000000000000000000060606066006062e°°°e°6°6 9 0000000000000000000000000000000000000000000000000009090909060060060000000000000000000606060606060606606606600°°°°6°6 900000000000000000000009000000000000000600000000000009090090000000000000000000000600000060606060°°eeeeee°66 ĐÔ ÔаĐÔ°ÔeÔ°°eÔC°GCÔC°°bÔeÔeÔbCÔe°o°od°0ôe0ôeôo°od°oô°eôoôeoôo°oô°oôôeoôoôoôeoeoo°o°o°oôoô°eeoeoôoôôeeoo°oôoôoô°eoeeo°o°oôo°eoeo°oôoôe°edeeee°o°oôeô°eeo°oôeeeee°eoeeeoeee°ee°e©°e© *0000000000000000000000000000000000006000060606060000000000000606060606060060600606000000000000000606060606066066°°°°e°6°e6 Ho Chi Minh city,. THESIS REVIEWER ACKNOWLEDGMENT Eirst of all, I would like to express my sincerest thanks to Dr. Do Trong Hop, Dr.
Tran Van Thanh who has enthusiastically guided and supported me throughout the process of studying and researching to complete the thesis. Besides teaching and commenting on academic knowledge, presentation skills, programs, research, reports,.The teacher also cares about students’ health and psychological status and always listens, shares, inspires, and motivates me to complete this thesis. The knowledge and skills that he has imparted will definitely be a piece of valuable baggage for my future growth. Additionally, I would like to thank the Department of Information Systems, Faculty of Information Science and Engineering, University of Information Technology — National University Ho Chi Minh City and all the teachers who have provided valuable knowledge during the past five years.
At the same time, I would also like to thank our defense committee, who commented on and provided expertise on our thesis. Lastly, I would like to thank Dr. Do Trong Hop, Dr. Tran Van Thanh and my family, who have supported and encouraged me to complete this thesis.
Nguyen Van Tuan TABLE OF CONTENTS o3LUex> THESIS ADVISOR'S COMMIENTTS. LH HH HH HH HH HH gu 4 THESIS REVIEWER 'S COMMENTTS. 6 TABLE OF CONTTENTTS.-- HH HH TH ng HH HH HH re 7 LIST OF FIGURES 17. 11 LIST OF TABLES 1177.
15 LIST OF ABBREVIA TIONS.- LH HH ng HH HH HH, 16 0Ý. 2 Problem sfaf€IT€TIL .- --- Gà TH ng ng nh ng nh ng ng 2 9001121. c5 c1 21111 TH ng HH HH.-- 5L 1 HT TH TH HH HH HH 6 RESUItS 1. HH HH HH HH ng Hư 13 2.
Research situation in the WOTÏ. Researches 1n OUT COUMULY .-- «+ s11 1E nh ng ng ệt 15 "6Šan. General information about the dataset. Specific information Of Objects.
Hydroelectric power DÏATIẲ.Ln SH TH TH ng TH HH HH HH Hệ 30 3. Wind power plant 7 .90 9000) cRIỌađđadđaaiaaa. sn tO aa Fe V. Water TS€TVOIT.
Ăn HH HH TH TH HH ng 40 3. Evaluation of collected data S€fS. 6 SG ng ng niên 42 3. Evaluation of data sets collected from SOULCES.
Essential in building new datasets. Method to create a dataset. Basis for building afAS€(.-- óc <1 vn ng ng ệt 43 3. Process for building new dafaS€fS.
Label the afa. Statistics for groups Of (4fA.- -- Ă ST TH TH TH TH ng ng rệt 47 4. Q00 TT TK 47 Chapter 4.- G1 HH HH krưy 51 4. - G- G11 HH HH Hi HH 53 4.
Identification within data range. Identification outside the data range .- - --ccc St rreg 54 4. EXPERIMENT AND EVALUATION. Comparison each data and Performance of modeÌs.
Metrics of each Class. Evaluate of each imO(el .- --‹- c s11 +3 191119119 1191 vn HH it 60 5. rrugmmms eee ee gggmmmmes. CNN combine with Pencil Sketch.
SVM combine with Pencil Sketch. Visual experiment results 1. CNN and Pencil SkefCH.- - s1 191 1 vn ng nến 84 h0 600. HH HT TH TH HH HH HT TH HH rt 110 5.
SVM and Pencil Sketch. Summary of experiment results. CONCLUSION AND FUTURE WORK. 140 LIST OF FIGURES Figure 0.1: Global Infrastructure Rankings 2011-2012 by World Economic Forum.2: ASEAN countries ranking in infrastructure 2017-2018.2: Map of airports in Viet Nam.3: Hydroelectric power plant.4: Map of airports in Viet Nam.6: Vietnam industrial zone's map.8: Vietnam stadium map for V-League.9: Water treatment plant.10 : List of water treatment plant.12: Bridges of Viet Nam.13: Wind power plant.14: Wind farm power plants in Viet Nam.16: Map of technology parks in Viet Nam.18: Students gather in shape of Vietnam map.20: List of water reservoir in Vietnam.22: List of water reservoir in Vietnam.23: Considering two types of infrastructure Highway and Hospital.24: Considering two types of infrastructure Bridge and Highway.1: Viet Nam ranking about infrastructure in the world.2: Diagram of idea.1: Performance Comparison of a CNN Model Across Different Infrastructure Sectors.2: Performance of CNN Model Using Pencil Sketch for Feature Extraction.3: RESNET-50 Model Performance Across Infrastructure Sectors.4: Infrastructure Classification Performance of VGG-16 Model.5: Infrastructure Classification Efficacy of SVM with Pencil Sketch Features Extracted.6: CNN Model predict Airport satellite image.7: CNN Model predict Bridge satellite image.8: CNN Model predict Expressway satellite image.9: CNN Model predict Hydroelectric power plant satellite image.10: CNN Model predict Wind power plant satellite image.11: CNN Model predict Water treatment plant satellite image.12: CNN Model predict Water reservoir satellite image.13: CNN Model predict Stadium satellite image.14: CNN Model predict School satellite image.15: CNN Model predict Industrial zone satellite image.16: CNN Model predict Technology park satellite image.17: Pencil Sketch and CNN Model predict Airport satellite image.18: Pencil Sketch and CNN Model predict Bridge satellite image.19: Pencil Sketch and CNN Model predict Expressway satellite image.20: Pencil Sketch and CNN Model predict Hydroelectric power plant satellite image.21: Pencil Sketch and CNN Model predict Wind power plant satellite image.22: Pencil Sketch and CNN Model predict Water treatment plant satellite image.23: Pencil Sketch and CNN Model predict Water reservoir satellite image.24: Pencil Sketch and CNN Model predict Stadium satellite image.25: Pencil Sketch and CNN Model predict School satellite image.26: Pencil Sketch and CNN Model predict Industrial zone satellite image.27: Pencil Sketch and CNN Model predict Technology park satellite image.28: Resnet50 Model predict Airport satellite image.29: Resnet50 Model predict Bridge satellite image.30: Resnet50 Model predict Expressway satellite image.31: Resnet50 Model predict Hydroelectric power plant satellite image.32: Resnet50 Model predict Wind power plant satellite image.33: Resnet50 Model predict Water treatment plant satellite image.34: Resnet50 Model predict Water reservoir satellite image.35: Resnet50 Model predict Stadium satellite image.36: Resnet50 Model predict School satellite image.37: Resnet50 Model predict Industrial zone satellite image.38: Resnet50 Model predict Technology park satellite image.39: VGG16 Model predict Airport satellite image.40: VGG16 Model predict Bridge satellite image.41: VGG16 Model predict Expressway satellite image.42: VGG16 Model predict Hydroelectric power plant satellite image.43: VGG16 Model predict Wind power plant satellite image.44: VGG16 Model predict Water treatment plant satellite image.45: VGG16 Model predict Water reservoir satellite image.46: VGG16 Model predict Stadium satellite image.47: VGG16 Model predict School satellite image.48: VGG16 Model predict Industrial zone satellite image.49: VGG16 Model predict Technology park satellite image.50: SVM Model and Pencil Sketch Method predict Airport satellite image.51: SVM Model and Pencil Sketch Method predict Bridge satellite image.52: SVM Model and Pencil Sketch Method predict Expressway satellite image.53: SVM Model and Pencil Sketch Method predict Hydroelectric power plant satellite image.54: SVM Model and Pencil Sketch Method predict Wind power plant satellite image.55: SVM Model and Pencil Sketch Method predict Water treatment plant satellite image.56: SVM Model and Pencil Sketch Method predict Water reservoir satellite image.57: SVM Model and Pencil Sketch Method predict Stadium satellite image.58: SVM Model and Pencil Sketch Method predict School satellite image.59: SVM Model and Pencil Sketch Method predict Industrial zone satellite image.60: SVM Model and Pencil Sketch Method predict Technology park satellite image.
135 LIST OE TABLES Table 3.1: Data from 3 selected types for crOSS-refer€nCe.-- ----«++«sss+sc++ 44 Table 3.2: Example to label the data. eee cee 2< E931 2111 1 9 1 1 trên 46 Table 3.-- s5 1H HH He 47 Table 3.4: Data for each infrastructure fYD€S.- - --c c SH HH rrg 48 Table 3.5: Data for each transportation fY€S.6: Data for each energy fYD€S.-- ---càng HH tưệt 49 Table 3.7: Data for each water ẨYDS.- nh TH TH HH nàn Hàng gi 49 Table 3.8: Data for each public fyD€S.- ---- c 1S SH ng 50 Table 3.9: Data for each production and industrial types.1: Table of Measurement results of the models for Transportation.2: Table of Measurement results of the models for Energy.3: Table of Measurement results of the models for Water.4: Table of Measurement results of the models for Public.5: Table of Measurement results of the models for Production and Industrial.6: CNN Model Image Classification Results for Infrastructure Types.7: CNN Model and Pencil Sketch Method Classification Results for Infrastructure [DafaSGK.- HT T ng nHggHH HH HHg 85 Table 5.8: ResNet50 Model Accuracy for Infrastructure Image Classification.9: VGG16 Model Image Classification Performance on Infrastructure Dataset.10: SVM Classifier Performance with Pencil Sketch Feature Extraction on Infrastructure ÏIAØ€S.- Ăn TH HH TH HH HH nh 124 LIST OF ABBREVIATIONS Ordinal | Abbreviations Meaning numbers 1 CNN Convolutional Neural Network 2 Resnet50 Residual Network 50 layers 3 VGG16 Visual Graphics Group 13 convolutional layers & 3 fully connected layers 4 SVM Support Vector Machine 5 KNN K-Nearest Neighbors 6 LR Logistic Regression 6 SIFT Scale-Invariant Feature Transform 7 HOG Histogram of Oriented Gradients ABSTRACT Image classification is one of the important topics in the field of image processing. This topic has been attracting much attention from researchers. In today's modern life, with the strong development of the Internet, especially the integration for comprehensive development in Vietnam, the amount of data increases, creating potential resources.
capabilities need to be exploited, analyzed, and expanded. This issue has attracted a lot of attention around the world, especially projects that affect the entire development of that country, but in Vietnam there are still few studies on this issue. Therefore, the goal of this thesis is to classify satellite images collected from Apple Maps, Google Maps, and the remote sensing database of the Vietnam Academy of Science and Technology on infrastructure. We perform a number of data preprocessing methods, apply current typical deep learning methods for image classification such as CNN, Resnet, VGG, and implement popular machine learning models such as Decision Tree, SVM, KNN, Logistic Regression and traditional feature extraction methods such as SIFT, HOG, especially with the Pencil Sketch image processing technique to smooth and highlight lines in the image.
At the same time, after installing individual models, we select the best machine learning model, combine it with the most suitable extraction method, and then combine them together to take advantage of the advantages of each model. model and evaluate the performance of image classification problem. Through our experiments, we find the most suitable processing methods, models, and techniques for the classification problem with the dataset we collected. We have achieved positive results for a new data set, applied to the specific region of Vietnam.
PREAMBLE s* Problem statement Classification refers to the process of categorizing items or data into predefined groups or classes based on their characteristics. In the context of machine learning and data analysis, classification involves training a model to recognize patterns and assign labels to input data. It is a type of supervised learning where the algorithm learns from labeled training data and then applies that learning to new, unseen data. Infrastructure generally refers to the basic physical and organizational structures and facilities needed for the operation of a society, enterprise, or system.
A country's infrastructure is the set of physical and organizational facilities necessary for the economy and society to function effectively. Infrastructure includes physical components such as roads, bridges, transportation systems, energy facilities, water supply and wastewater treatment, as well as living space structures such as housing , schools, and medical facilities.