Preprocessed Mask Region-Based Convolutional Neural Network Untuk Deteksi Tempat Parkir Mobil Pada Sistem Parkir Cerdas

Naufal, Ahmad Afiif (2020) Preprocessed Mask Region-Based Convolutional Neural Network Untuk Deteksi Tempat Parkir Mobil Pada Sistem Parkir Cerdas. Masters thesis, Institut Teknologi Sepuluh Nopember.

[thumbnail of 05111850010023-Master_Thesis.pdf]
Preview
Text
05111850010023-Master_Thesis.pdf

Download (8MB) | Preview

Abstract

Pengemudi mobil di kota-kota besar seringkali kesulitan dalam menemukan tempat parkir karena terbatasnya lahan parkir dan banyaknya jumlah mobil. Sistem parkir cerdas diusulkan untuk memudahkan pencarian tempat parkir. Deteksi mobil dalam pendekatan tradisional menggunakan beberapa sensor, dimana setiap satu tempat parkir mobil harus memiliki satu sensor. Oleh karenanya, dibutuhkan sensor dalam jumlah besar untuk mengatasi area parkir yang memiliki tempat parkir dengan jumlah besar sehingga memakan banyak biaya.
Sebuah sistem diusulkan untuk mengurangi biaya tersebut, di mana algoritma deteksi mobil otomatis berbasis kamera CCTV dikembangkan untuk menggantikan sensor. Penelitian ini mengusulkan Preprocessed Mask Regionbased Convolutional Neural Network (Mask R-CNN) untuk deteksi tempat parkir secara otomatis. Deteksi tempat parkir pada area parkir terbuka dapat gagal karena adanya perubahan kondisi cahaya. Tahap praproses dengan mengkombinasikan peningkatan kontras menggunakan kerangka kerja Exposure Fusion bertujuan mengatasi masalah tersebut.
Hasil dari metode yang diusulkan dapat meningkatkan deteksi tempat parkir mobil bila dibandingkan tanpa praproses. Performa deteksi tempat parkir mobil mencapai akurasi IoU sebesar 85.80%. Pada metode klasifikasi didapatkan hasil akurasi sistem deteksi ketersediaan tempat parkir mobil terbaik sebesar 73.73%.
==================================================================================================================
Car drivers in big cities often encounter difficulty in finding a parking spaces because of limited parking spaces and a large number of cars. To solve the problem, a smart parking system being proposed to ease the process of searching for parking spaces. Detection of the car in the traditional approach using some sensors, in which each parking space must have one sensor on it. Therefore, a large number of sensors are needed in a larger parking area and it will cost a lot.
A system was proposed to reduce the costs, in which an automatic car detection algorithm based CCTV is developed as a substitute for the sensor detection. This research proposed a Preprocessed Mask Region-based Convolutional Neural Network (Mask R-CNN) to detect parking spaces automatically. However, detecting a parking spaces in open parking areas can led to failure due to changes in light conditions. The preprocessing method was introduced by applying contrast enhancement of the Fusion Exposure framework to solve the problem.
The results of the proposed method can improve the detection of car parking spaces compared to without preprocessing. Parking space detection performance reached an Intersection over Union (IoU) accuracy of 85.80%. On the other hand, the classification method for parking space status detection obtained the best accuracy of 73.73%.

Item Type: Thesis (Masters)
Uncontrolled Keywords: Smart Parking Systems, Car Parking Spaces Detection, Convolutional Neural Network, Mask R-CNN, Image Enhancement, Sistem Parkir Cerdas, Deteksi Tempat Parkir Mobil, Convolutional Neural Network, Mask R-CNN, Peningkatan Citra
Subjects: Q Science > QA Mathematics > QA76.6 Computer programming.
Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
Q Science > QA Mathematics > QA76.9.D33 Data compression (Computer science)
Divisions: Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55101-(S2) Master Thesis
Depositing User: Ahmad Afiif Naufal
Date Deposited: 16 Aug 2020 01:48
Last Modified: 20 Jun 2023 13:46
URI: http://repository.its.ac.id/id/eprint/78222

Actions (login required)

View Item View Item