PENDETEKSI MANUSIA MENGGUNAKAN HISTOGRAM OF ORIENTED

GRADIENT (HOG) BERBASIS JETSON TK1 DAN STREAMING VIA REAL-TIME

STREAMING PROTOCOL (RTSP)

oleh

Tectona Dewa Putri

NIM: 622013005

Skripsi

Untuk melengkapi salah satu syarat memperoleh

Gelar Sarjana Teknik

Program Studi Teknik Komputer

Fakultas Teknik Elektronika dan Komputer

Universitas Kristen Satya Wacana

Salatiga

Mei 2019

ii

iii

iv

iii

v

KATA PENGANTAR

Puji dan syukur penulis ucapkan kepada Allah SWT yang atas limpahan berkat dan

rahmat-Nya penulis mampu menyelesaikan penulisan tugas akhir sebagai syarat kelulusan di

Fakultas Teknik Elektronika dan Komputer Universitas Kristen Satya Wacana. Pada

kesempatan ini penulis juga hendak mengucapkan terima kasih kepada berbagai pihak yang

baik secara langsung maupun tidak telah membantu penulis dalam menyelesaikan skripsi ini :

1. Bapak Banu Wirawan Yohanes M.Comp.Sc. dan Bapak Hartanto Kusuma W, M.T

selaku pembimbing atas bimbingan, saran serta nasehat yang telah diberikan selama

proses pembuatan skripsi.

2. Bapak Atyanta Nika Rukmasari, S.T,M.E selaku wali studi atas arahan serta nasehat

yang telah diberikan selama berkuliah di Fakultas Elektronika dan Komputer

Universitas Kristen Satya Wacana.

3. Seluruh staff dosen atas ilmu yang telah diberikan selama berkuliah di Universitas

Kristen Satya Wacana.

4. Kedua orang tua terkasih dan adik saya Switenia Dewa Putri atas dukungan doa,

moral, serta materiil yang telah diberikan kepada penulis.

5. Keluarga besar angkatan 2013 sebagai teman seperjuangan yang selalu memberi

dukungan kepada penulis.

6. Seluruh karyawan Tata Usaha yang telah membantu kelancaran penyelesaian

pembuatan skripsi ini.

7. Seluruh pihak yang membantu penulis selama proses pembuatan skripsi yang tidak

dapat disebutkan satu per satu.

Dalam penulisan skripsi ini penulis menyadari masih banyak kekurangan baik dalam

isi, penyampaian, serta penulisan sehingga dibutuhkan kritik dan saran yang membangun

demi perbaikan skripsi ini ke depannya.

Salatiga, Mei 2019

Penulis,

Tectona Dewa Putri

vi

vii

ABSTRACT

The detection process in image processing is an interesting topic and has been

developed by many researchers, both those applied to robotics systems and in the security

field. In the field of security the use of the camera is realized as a visual sensor for detecting

objects and calculating the number of queues. At this time the camera that is often used is

Closed Circuit Television (CCTV) cameras, CCTV cameras are able to identify all objects

captured by the camera, at the same time the results of CCTV camera detection can be

monitored directly through other computer devices.

This thesis contains the design of a human object detection system and can stream

videos. One method used to detect the presence of human objects is feature extraction using

the Histogram of Oriented Gradient (HOG) method that takes the edge or gradient structure

that is characterized by local or edge direction with a good local gradient intensity

distribution and uses the Support Vector Machine (SVM) classification as separating two

classes in two input spaces and for streaming services will use VLC software with the Real-

Time Streaming Protocol (RTSP) protocol that functions as communication between the

server and the client.

Implementation of the system produces the best parameter is γ = 0.000001 and C =

1000 in the training process and the accuracy of the testing reached 97.59%, as well as

running a human object detection system that streams with the help of VLC app uses the

RTSP protocol with ± 3 second delay.

viii

DAFTAR ISI

INTISARI ............................................................................................................................... i

ABSTRACT .......................................................................................................................... ii

KATA PENGANTAR........................................................................................................... iii

DAFTAR ISI ........................................................................................................................ iv

DAFTAR GAMBAR ............................................................................................................ vi

DAFTAR TABEL ................................................................................................................ vii

DAFTAR KODE ................................................................................................................ viii

DAFTAR RUMUS ............................................................................................................... ix

BAB I PENDAHULUAN ...................................................................................................... 1

1.1 Tujuan ........................................................................................................................ 1

1.2 Latar Belakang ........................................................................................................... 1

1.3 Gambaran Sistem ....................................................................................................... 2

1.4 Spesifikasi Sistem ....................................................................................................... 2

1.5 Sistematika Penulisan ................................................................................................. 3

BAB II DASAR TEORI ........................................................................................................ 4

2.1 Dataset ....................................................................................................................... 4

2.2 Histogram of Oriented Gradient (HOG) ..................................................................... 4

2.2.1 Konversi Citra atau Normalisasi Warna........................................................... 5

2.2.2 Menghitung Nilai Graident Citra ..................................................................... 5

2.2.3 Menentukan Bin Orientasi ............................................................................... 6

2.2.4 Normalisasi Blok ............................................................................................ 7

2.2.5 Deskriptor Windows ........................................................................................ 7

2.2.6 Principal Component Analysis ........................................................................ 8

2.2.7 Klasifikasi Support Vector Machine (SVM) .................................................. 10

2.3 Metode Pengujian ..................................................................................................... 15

ix

2.3.1 Train and Test Split ....................................................................................... 15

2.3.2 K-Fold Cross Validation ............................................................................... 15

2.3.3 Confusion Matrix .......................................................................................... 16

2.4 Real-Time Streaming Protocol (RTSP) ..................................................................... 17

2.5 Nvidia Jeston TK1 .................................................................................................... 18

BAB III PERANCANGAN DAN IMPLEMENTASI ........................................................... 20

3.1 Perancangan Sistem Server ....................................................................................... 20

3.1.1 Instalasi OpenCV dan Python pada Jetson TK1 ........................................... 21

3.1.2 Input Dataset ............................................................................................... 22

3.1.3 Ekstraksi Fitur ............................................................................................. 22

3.1.4 Principal Component Analysis .................................................................... 23

3.1.5 Learning SVM ............................................................................................ 24

3.1.6 Deteksi Obyek Manusia .............................................................................. 25

3.1.7 Metode Pengujian ....................................................................................... 27

3.1.8 Pengiriman Paket data Streaming ................................................................ 28

3.2 Perancangan Sistem Client ....................................................................................... 32

BAB IV PENGUJIAN DAN ANALISIS .............................................................................. 34

4.1 Penggunaan Data Uji ................................................................................................ 35

4.2 Pengujian Dataset dan Pencarian Parameter .............................................................. 35

4.3 Deteksi Obyek Manusia ............................................................................................ 44

4.4 Pengujian Streaming menggunakan Protokol RTSP .................................................. 45

4.5 Analisis Hasil Pengujian ........................................................................................... 46

BAB V KESIMPULAN DAN SARAN ................................................................................ 49

5.1 Kesimpulan .............................................................................................................. 49

5.2 Saran ....................................................................................................................... 49

DAFTAR PUSTAKA .......................................................................................................... 50

x

DAFTAR GAMBAR

Gambar 2.1. Tahapan metode HOG-PCA [1] .....................................................................5

Gambar 2.2 Contoh Penggunaan PCA [3] ..........................................................................7

Gambar 2.3 Fitur HOG diekstraksi dari semua lokasi [2] ....................................................9

Gambar 2.3 Poin yang dipilih oleh PCA [2] ......................................................................9

Gambar 2.4. SVM berusaha menemukan hyperline terbaik[4] .......................................... 10

Gambar 2.5 Fungsi Φ memetakan data [4] ............................................. .................. 14

Gambar 2.6. Contoh Train and Test Split [6] .................................................................... 15

Gambar 2.7. Contoh K-Fold Cross Validation [7] ............................................................ 16

Gambar 3.1 Diagram Sistem ............................................................................................ 20

Gambar 3.2 Diagram blok proses learning SVM dan deteksi obyek manusia pada server . 21

Gambar 3.3 Flowchart proses ekstraksi fitur HOG ........................................................... 23

Gambar 3.4 Flowchart sederhana algoritma training SVM ............................................... 24

Gambar 3.5 Flowchart sederhana algoritma testing SVM ................................................ 24

Gambar 3.6 Flowchart proses pendeteksi obyek manusia menggunakan HOG ................ 26

Gambar 3.7. Flowchart sederhana algoritma Train and Test Split pada pengujian ............ 27

Gambar 3.8. Flowchart sederhana algoritma K-Fold Cross Validation ............................. 28

Gambar 3.9 Tampilan VLC .............................................................................................. 29

Gambar 4.1 Hasil deteksi obyek manusia ......................................................................... 44

Gambar 4.2 Stream menggunakan protokol RTSP ............................................................ 45

xi

DAFTAR TABEL

Tabel 2.1 Kernel yang umum dipakai dalam SVM [4] ..................................................... 14

Tabel 2.2 Klasifikasi binary dengan confusion matrix ....................................................... 17

Tabel 4.1 Pengujian kinerja menggunakan HOG-SVM kernel Linear – Train and Test Split

dan K-Fold Cross Validation pada dataset INRIA ............................................. 35

Tabel 4.2 Pengujian kinerja menggunakan HOG-SVM kernel RBF – Train and Test Split

pada dataset INRIA ........................................................................................... 36

Table 4.3 Pengujian kinerja menggunakan HOG-SVM kernel RBF – K-Fold Cross

Validation pada dataset INRIA .......................................................................... 37

Tabel 4.4 Pengujian kinerja menggunakan HOG-PCA-SVM kernel Linear – Train and Test

Split dan K-Fold Cross Validation pada dataset INRIA ..................................... 37

Tabel 4.5 Pengujian kinerja menggunakan HOG-PCA-SVM kernel RBF – Train and Test

Split pada dataset INRIA ................................................................................... 38

Tabel 4.6 Pengujian kinerja menggunakan HOG-PCA-SVM kernel RBF – K-Fold Cross

Validation pada dataset INRIA .......................................................................... 39

Tabel 4.7 Pengujian kinerja menggunakan HOG-SVM kernel Linear – Train and Test Split

dan K-Fold Cross Validation pada dataset BB-5 ................................................ 39

Tabel 4.8 Pengujian kinerja menggunakan HOG-SVM kernel RBF – Train and Test Split

pada dataset BB-5 ............................................................................................. 40

Tabel 4.9 Pengujian kinerja menggunakan HOG-SVM kernel RBF – K-Fold Cross

Validation pada dataset BB-5 ............................................................................ 41

Tabel 4.10 Pengujian kinerja menggunakan HOG-PCA-SVM kernel Linear – Train and Test

Split dan K-Fold Cross Validation pada dataset BB-5 ........................................ 41

Tabel 4.11 Pengujian kinerja menggunakan HOG-PCA-SVM kernel RBF – Train and Test

Split pada dataset BB-5 ..................................................................................... 42

Tabel 4.12 Pengujian kinerja menggunakan HOG-PCA-SVM kernel RBF – K-Fold Cross

Validation pada dataset BB-5 ............................................................................ 43

xii

DAFTAR KODE

Kode 3.1. Mengaktifkan repository universal ..................................................................... 21

Kode 3.2 Instalasi OpenCV ................................................................................................ 21

Kode 3.3 Instalasi Python ................................................................................................... 21

Kode 3.4 Pemanggilan fungsi input dataset citra training SVM .......................................... 22

Kode 3.5 Pemanggilan fungsi ektraksi fitur HOG ............................................................... 23

Kode 3.6 Inisialisasi SVM, pemanggilan fungsi train SVM dan fungsi prediksi SVM pada

bahasa python dengan library Scikit-learn ............................................................ 25

Kode 3.7 Implementasi dari HOG dan SVM pada library OpenCV .................................... 26

xiii

DAFTAR RUMUS

Rumus 2.1 Konversi citra RGB ............................................................................................ 5

Rumus 2.2 Menghitung gradient citra ................................................................................... 6

Rumus 2.3 Normalisasi blok ................................................................................................ 7

Rumus 2.4 Persamaan Difference of Gaussian (GOD) .......................................................... 9

Rumus 2.5 Menghitung fitur baru HOG-PCA ....................................................................... 9

Rumus 2.6 Menghitung vektor pelatihan PCA ...................................................................... 9

Rumus 2.6 Persamaan Quadratic Programming (QP) ........................................................ 11

Rumus 2.7 Persamaan Lagrange Multiplier ........................................................................ 12

Rumus 2.8 Optimalisasi parameter SVM ............................................................................ 12

Rumus 2.9 Fungsi Kernel Trick pada SVM ........................................................................ 14

Rumus 2.10 Prediksi Klasifikasi SVM ............................................................................... 15

Rumus 2.11 Rumus perhitungan Akurasi, Presisi dan Recall .............................................. 17