Sistem Navigasi pada Mobil Cerdas Menggunakan Lane Detection dan Segmentasi Semantik

Satriojati, Ikhwan Hakim (2023) Sistem Navigasi pada Mobil Cerdas Menggunakan Lane Detection dan Segmentasi Semantik. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 07111940000071-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2025. Download (5MB) | Request a copy

Abstract

Salah satu fitur penting dalam persepsi mobil otonom adalah kemampuannya untuk mendeteksi jalur (lane detection) dan bergerak sesuai jalur tersebut. Dengan menggunakan visi komputer, gambar yang ditangkap kamera dapat diproses sehingga dapat mendeteksi jalur bagi kendaraan. Namun karena ketidaksamaan karakteristik pada setiap jalan dan kondisi pencahayaan yang berubah membuat hal ini semankin sulit bagi visi komputer sehingga perlu digunakan solusi yang lebih robust untuk melakukan lane detection dengan mengkombinasikan visi komputer dengan deep learning yaitu segmentasi semantik (semantic segmentation). Pada tugas akhir ini, sistem lane detection akan dibuat dengan menggunakan segmentasi semantik untuk mensegmentasi jalan dari gambar yang diambil oleh kamera yang terpasang pada mobil. Kemudian hasil output dari segmentasi akan dijadikan sebagai input untuk algoritma kontrol mobil sehingga mobil nantinya akan mampu bergerak mengikuti kontur jalan. Sistem dibangun menggunakan framework ROS (Robot Operating System). Dataset yang digunakan didapatkan dari gambar jalan yang diambil di lingkungan Institut Teknologi Sepuluh Nopember. Model segmentasi semantik yang digunakan adalah ENet. Training dilakukan dengan menggunakan optimasi Adam dan fungsi loss cross entropy dengan menggunakan framework PyTorch. Dari hasil training didapatkan model dengan nilai mIoU (mean Intersection over Union) sebesar 0.9834. Dari hasil pengujian, mobil mampu bergerak mengikuti kontur jalan dan sistem mampu berjalan pada rata-rata 10 fps keatas.
===================================================================================================================================
One of the important features in the perception of autonomous vehicles is their ability to detect lanes and move accordingly. By using computer vision, images captured by cameras can be processed to detect lanes for vehicles. However, the variations in characteristics on different roads and changing lighting conditions make it increasingly challenging for computer vision alone. Therefore, a more robust solution is needed to perform lane detection by combining computer vision with deep learning, specifically semantic segmentation. In this final project, a lane detection system will be developed using semantic segmentation to segment the road from the images taken by the camera installed on the vehicle. The output from the segmentation will then be used as input for the vehicle control algorithm, enabling the vehicle to move along the road contours. The system is built using the ROS (Robot Operating System) framework. The dataset used for training consists of road images captured in the environment of the Sepuluh Nopember Institute of Technology. The ENet model is employed for semantic segmentation. Training is performed using the Adam optimization algorithm and cross-entropy loss function within the PyTorch framework. The trained model achieves an mIoU (mean Intersection over Union) value of 0.9834. From the testing results, the vehicle can follow the road contours, and the system operates at an aveage frame rate of over 10 fps.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Computer vision, Deep learning, Lane detection, Robot Operating System Semantic segmentation.
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7882.P3 Pattern recognition systems T Technology > TL Motor vehicles. Aeronautics. Astronautics > TL152.8 Vehicles, Remotely piloted. Autonomous vehicles.
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Electrical Engineering > 20201-(S1) Undergraduate Thesis
Depositing User:	Ikhwan Hakim Satriojati
Date Deposited:	28 Jul 2023 15:20
Last Modified:	28 Jul 2023 15:20
URI:	http://repository.its.ac.id/id/eprint/99616

Actions (login required)

View Item