Detecting Incoming and Outgoing Passengers on Intelligent Car (iCar Its) Using Computer Vision

Putri, Vania Huwaida (2023) Detecting Incoming and Outgoing Passengers on Intelligent Car (iCar Its) Using Computer Vision. Other thesis, Institut Teknologi Sepuluh Nopember.

[thumbnail of 07111942000005-Undergraduate_Thesis.pdf] Text
07111942000005-Undergraduate_Thesis.pdf - Accepted Version
Restricted to Repository staff only until 1 September 2025.

Download (4MB) | Request a copy

Abstract

Intelligent Car atau disebut iCar ITS merupakan sebuah transportasi umum di lingkungan kampus ITS, yang menggabungkan tenaga listrik, otonomi, Artificial Intelligence (AI), dan Internet of Things (IoT). Tujuan dari proyek akhir ini adalah untuk menggunakan teknologi visi komputer untuk mengidentifikasi dan menghitung jumlah penumpang yang masuk dan keluar. Untuk mencapainya, metode You Only Look Once (YOLO) dengan fokus khusus pada deteksi manusia. YOLOv5 dari Ultralytics digunakan untuk proyek akhir ini. YOLOv5 dari Ultralytics digunakan untuk proyek akhir ini, algoritma tersebut terintegrasi dengan ekosistem PyTorch, yang menyederhanakan implementasi, membuat environment lebih mudah dibandingkan dengan environment YOLO sebelumnya. Ada tiga bagian penting dalam arsitektur YOLOv5, yaitu Backbone, Neck, dan Head. Backbone terdiri dari beberapa iterasi Cross-Stage Partial (CSP) atau CSPNet, bagian Neck menggunakan PANet untuk mencapai fitur piramida, dan bagian Head adalah bagian terakhir yang bertanggung jawab atas prediksi bounding box dan probabilitas kelas objek. Selain itu, YOLOv5 digunakan untuk menghemat waktu komputasi dan mengurangi beban kerja pada komputer, karena lebih ringan dibandingkan dengan YOLOv7 atau YOLOv8. Data pelatihan dan validasi untuk model kustom diperoleh dari database iCar, yang terdiri dari gambar dan video yang diambil selama operasi iCar di kampus. Implementasi sistem hitung menggunakan PyTorch dan OpenCV untuk mengembangkan sistem penghitung dan menjalankan model YOLO. Sistem ini juga menggunakan ID yang disimpan untuk melacak penumpang yang terdeteksi. Sistem penghitung penumpang yang masuk menunjukkan kinerja akurasi rata-rata sebesar 97,5%, yang dicapai setelah penyesuaian confidence score. Pengujian sistem data skenario memperoleh akurasi 100% dalam menghitung penumpang yang datang dan pergi. Hasil ini menunjukkan bahwa YOLOv5 cocok untuk sistem deteksi dan penghitungan pada penumpang iCar
=====================================================================================================================================
The Intelligent Car (iCar ITS) represents an innovative approach to public transportation, combining electric power, autonomy, Artificial Intelligence (AI), and the Internet of Things (IoT). The aim of this final project is to use computer vision technology to identify and count incoming and outgoing passengers. To achieve this, the You Only Look Once (YOLO) method is employed, specifically focusing on human detection. YOLOv5 from Ultralytics is used for the final project. It was integrated with the PyTorch ecosystem, simplifying deployments, and making the environment easier to work with than previous YOLO environments. There are three essential parts to the YOLOv5 architecture, namely the Backbone, Neck, and Head. The Backbone comprises multiple iterations of the Cross-Stage Partial Network (CSPNet), the Neck section utilizes PANet to achieve pyramid features, and the head is responsible for predicting bounding boxes and object class probabilities. Furthermore, YOLOv5 is used to save computation time and reduce the workload on the computer, as it is lighter than YOLOv7 or YOLOv8. The training and validation data for the custom model were obtained from the iCar database, consisting of images and videos captured during the iCar's operations on campus. The counter system was implemented using PyTorch and OpenCV to develop the counting mechanism and execute the YOLO model. It also utilizes the stored ID to keep track of detected passengers. The incoming passenger counter system achieved an average accuracy performance of 97.5% after adjusting the confidence score. Scenario data system testing obtained 100% accuracy in counting incoming and outgoing passengers. These results show that YOLOv5 is suitable for the detection and counting system.

Item Type: Thesis (Other)
Uncontrolled Keywords: Computer Vision, Counter System, Human Detection, iCar ITS, YOLOv5, Human Detection, iCar ITS, Sistem Penghitung, Visi Komputer, YOLOv5.
Subjects: Q Science > Q Science (General) > Q325.5 Machine learning.
Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
T Technology > TA Engineering (General). Civil engineering (General) > TA1637 Image processing--Digital techniques
Divisions: Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Electrical Engineering > 20201-(S1) Undergraduate Thesis
Depositing User: Vania Huwaida Putri
Date Deposited: 02 Aug 2023 06:44
Last Modified: 02 Aug 2023 06:44
URI: http://repository.its.ac.id/id/eprint/100365

Actions (login required)

View Item View Item