Pengembangan Perangkat Lunak Berbasis Ai Deteksi Pedestrian Untuk Fasilitas Pedestrian

Herjuna, Arde Dewantara (2026) Pengembangan Perangkat Lunak Berbasis Ai Deteksi Pedestrian Untuk Fasilitas Pedestrian. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6012241027-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (5MB) | Request a copy

Abstract

Perkembangan urbanisasi yang pesat menuntut perencanaan infrastruktur pedestrian yang akurat dan responsif. Metode survei manual yang selama ini digunakan memiliki keterbatasan dalam hal akurasi, waktu, dan kemampuan penyediaan data real-time. Penelitian ini bertujuan mengembangkan sistem deteksi dan analisis perilaku pedestrian berbasis kecerdasan buatan (AI) dengan memanfaatkan teknologi deep learning model You Only Look Once (YOLO). Dataset dikumpulkan melalui rekaman video udara menggunakan drone di koridor Jalan Gubernur Suryo dan Jalan Jenderal Basuki Rahmat, Surabaya, kemudian diolah dengan teknik dynamic tiling dan dianotasi untuk klasifikasi demografis jenis kelamin dan kelompok usia. Model dilatih selama 100 epoch pada platform Google Colab dan dievaluasi menggunakan metrik mean Average Precision (mAP). Hasil penelitian menunjukkan model mencapai mAP@0.5 sebesar 0,821 dengan tingkat akurasi deteksi 94,29% (error 5,71%). Model berhasil mengekstraksi parameter kinerja pedestrian seperti kecepatan rata-rata 1,07 m/s, kepadatan 0,1545 orang/m², ruang pedestrian 6,4733 m²/orang, serta laju aliran 0,62 ped/menit/m pada sisi kiri dan 2,30 ped/menit/m pada sisi kanan, yang mengindikasikan Level of Service (LoS) kategori A yang berfungsi sebagai indikator kebutuhan perluasan trotoar jika terjadi penurunan tingkat pelayanan. Selain itu, model mampu menghasilkan Matriks Asal-Tujuan (Origin-Destination Matrix) dan visualisasi desire lines sebagai dasar empiris untuk penempatan crossing, dan integrasi dengan gravity model untuk peramalan pergerakan. Sistem yang dikembangkan telah diimplementasikan dalam bentuk aplikasi executable (.exe) untuk evaluasi fasilitas pedestrian secara real-time.
======================================================================================================================================
The rapid pace of urbanization demands accurate and responsive pedestrian infrastructure planning. Conventional manual survey methods have limitations in terms of accuracy, time efficiency, and the ability to provide real-time data. This study aims to develop an AI-based pedestrian detection and behavioral analysis system using the You Only Look Once (YOLO) deep learning model. The dataset was obtained from aerial video footage captured by drones along the pedestrian corridors of Jalan Gubernur Suryo and Jalan Jenderal Basuki Rahmat in Surabaya, then processed using dynamic tiling and annotated for demographic classification based on gender and age groups. The model was trained for 100 epochs on Google Colab and evaluated using mean Average Precision (mAP). The results show that the model achieved an mAP@0.5 of 0.821 with a detection accuracy of 94.29% (5.71% error). The model successfully extracted key pedestrian performance parameters, including an average walking speed of 1.07 m/s, density of 0.1545 persons/m², pedestrian space of 6.4733 m²/person, and pedestrian flow rates of 0.62 ped/min/m on the left side and 2.30 ped/min/m on the right side, indicating a Level of Service (LoS) category A, which serves as an indicator for potential sidewalk expansion if service levels decline. In addition, the model produces an Origin–Destination Matrix and desire line visualizations to support evidence-based crossing placement and integration with a gravity model for movement forecasting. The developed system has been implemented as an executable (.exe) application to enable real-time pedestrian facility evaluation.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Deteksi Pejalan Kaki, Deep Learning, Kecerdasan Buatan, Perencanaan Fasilitas Pedestrian, YOLO.  Artificial Intelligence, Deep Learning, Pedestrian Detection, Pedestrian Facility Planning, YOLO.
Subjects:	T Technology > TA Engineering (General). Civil engineering (General) > TA1637 Image processing--Digital techniques. Image analysis--Data processing.
Divisions:	Faculty of Civil, Environmental, and Geo Engineering > Civil Engineering > 22101-(S2) Master Theses
Depositing User:	Arde Dewantara Herjuna
Date Deposited:	27 Jan 2026 03:12
Last Modified:	27 Jan 2026 03:12
URI:	http://repository.its.ac.id/id/eprint/130551

Actions (login required)

View Item