Laurentius Kuncoro Probo Saputra, Laurentius Kuncoro Probo
Universitas Kristen Duta Wacana

Published : 3 Documents
Articles

Found 3 Documents
Search

Perbandingan Varian Metode Multiscale Retinex untuk Peningkatan Akurasi Deteksi Wajah Adaboost HAAR-like Saputra, Laurentius Kuncoro Probo
Jurnal Teknik Informatika dan Sistem Informasi Vol 2 No 1 (2016): JuTISI
Publisher : Maranatha University Press

Show Abstract | Original Source | Check in Google Scholar | Full PDF (2366.194 KB)

Abstract

Face detection is a popular research in image processing field. Face detection can be used in many application like multi-face recognition, video surveillance, human counter or monitoring. The famous face detection method is developed by Viola-Jones that is named Adaboost using HAAR-like feature. In many research about face detection using Adaboost HAAR-like, it is shown Adaboost HAAR-like face detection method have a limitation in low illumination. In this paper, we want to compare an improvement for increasing accuracy of face detection result using MSRCR and AMSR method. MSRCR and AMSR is an image enhancement method. Finally the results show that MSRCR is better than AMSR for increasing accuracy of face detection result. MSRCR can improve the accuracy until 1,43 times, but AMSR can only improve the accuracy until 1,11 times.
Pemanfaatan Raspberry Pi untuk Sistem Penghitung Mobil Otomatis pada Kampus UKDW Nugraha, Kristian Adi; Saputra, Laurentius Kuncoro Probo
Jurnal Teknik Informatika dan Sistem Informasi Vol 3 No 3 (2017): JuTISI
Publisher : Maranatha University Press

Show Abstract | Original Source | Check in Google Scholar | Full PDF (1010.273 KB)

Abstract

The number of vehicle keep increasing over time, causing some problems such as traffic jam and requirement of extra parking space. Universitas Kristen Duta Wacana (UKDW) is an university that has similar problem, lack of parking space at certain hours. That won't be a problem if campus activities such as course, student activities, etc. can spread evenly at all hours, not focused at one time only. But determine the accurate schedule is a difficult task, because it requires someone who observes all vehicles entering and exiting campus area all over time. This research propose a solution to create automatic system that can count all vehicles (car) that entering and exiting campus area, then the data can be used as a consideration in determining the scheduling at the next time, the lack of parking space area can be avoided. The system was built with internet of things technology using Raspberry Pi and camera as the main component with main server at the backend to store the data. That system is also using background substraction algorithm for counting  the number of vehicles that entering and exiting campus area. The accuracy of the system at counting vehicles is 70%.
Perbandingan Metode Klasifikasi untuk Menentukan Tingkat Kenyamanan Suhu pada Kondisi Rileks Berbasis Sinyal EEG Saputra, Laurentius Kuncoro Probo; Ratri, Ignatia Dhian Estu Karisma
Ultimatics : Jurnal Teknik Informatika Vol 10 No 2 (2018): Ultimatics : Jurnal Teknik Informatika
Publisher : Program Studi Teknik Informatika UMN

Show Abstract | Original Source | Check in Google Scholar

Abstract

Temperature control on air conditioner devices is still oriented to the target environment. This control mode ignores one's physiological condition. A person's thermal comfort varies when indoors. Thermal comfort is closely related to environmental thermal satisfaction conditions. EEG signal is a signal that can reflect brain activity. This research objective is provide classifier model for classifiying person’s thermal comfort based on eeg signal. This research used three conditions of room’s temperature. The features used by classfier are avarage frequency band, HFD, PFD, and MSE features. Classifier performance was assessed using ROC curve evaluation. The results of the classification of thermal comfort levels with EEG signals with the KNN classifier are obtained only by using the band frequency average feature, which is equal to 0.878 with a standard deviation of 0.022. While the SVM classifier gets the highest performance by using a combination of the average band + HFD frequency feature, which is 0.877 with a standard deviation of 0.013 in the linear kernel and RBF.