{"title":"Substraksi Latar Menggunakan Nilai Mean Untuk Klasifikasi Kendaraan Bergerak Berbasis Deep Learning","authors":"Ilal Mahdi, Kahlil Muchtar, Fitri Arnia, Tiara Ernita","doi":"10.17529/jre.v18i2.25224","DOIUrl":null,"url":null,"abstract":"Abstrak —Sistem deteksi objek bergerak telah banyak digunakan dalam kehidupan sehari-hari. Saat ini penelitian dibidang subtraksi latar masih terus dilakukan untuk mencapai hasil akurasi yang maksimal. Penelitian ini bertujuan untuk memodelkan substraksi latar dari sebuah citra menggunakan nilai mean dengan konsep non overlapping block . Selanjutnya, hasil substraksi latar akan digunakan dalam deteksi objek bergerak berbasis deep learning . Secara spesifik, citra masukan akan dibagi menjadi beberapa blok, kemudian nilai mean dari setiap blok akan dihitung untuk nantinya menghasilkan blok biner ( binary map ). Blok biner yang telah dihasilkan akan dijadikan sebagai masukan pembangkitan model latar ( background modelling ). Model latar bertujuan untuk memisahkan objek bergerak dengan latar yang ada pada citra masukan. Objek bergerak yang dihasilkan (lokalisasi objek) akan dikirimkan ke tahap klasifikasi objek menggunakan deep learning . Dataset yang digunakan dalam penelitian ini adalah CDNet 2014. Hasil penelitian mampu menghasilkan sistem deteksi objek Abstract — Moving object detection systems have been widely used in everyday life. Currently, research in the field of background subtraction is still being carried out to achieve maximum accuracy results. This study aims to model the background subtraction of an image using the mean value with the concept of non overlapping block. Furthermore, the background abstraction results will be used in deep learning-based moving object detection. Specifically, the input image will be divided into several blocks, then the mean value of each block will be calculated to later produce a binary block (binary map). The binary blocks that have been generated will be used as input for background modeling. The background model aims to separate moving objects from the background in the input image. The resulting moving object (object localization) will be sent to the object classification stage using deep learning. The dataset used in this study is CDNet 2014. The results of the study were able to produce a more accurate moving object detection system. Quantitative tests carried out resulted in an accuracy of above 90%.","PeriodicalId":30766,"journal":{"name":"Jurnal Rekayasa Elektrika","volume":null,"pages":null},"PeriodicalIF":0.0000,"publicationDate":"2022-07-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Jurnal Rekayasa Elektrika","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.17529/jre.v18i2.25224","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
Abstrak —Sistem deteksi objek bergerak telah banyak digunakan dalam kehidupan sehari-hari. Saat ini penelitian dibidang subtraksi latar masih terus dilakukan untuk mencapai hasil akurasi yang maksimal. Penelitian ini bertujuan untuk memodelkan substraksi latar dari sebuah citra menggunakan nilai mean dengan konsep non overlapping block . Selanjutnya, hasil substraksi latar akan digunakan dalam deteksi objek bergerak berbasis deep learning . Secara spesifik, citra masukan akan dibagi menjadi beberapa blok, kemudian nilai mean dari setiap blok akan dihitung untuk nantinya menghasilkan blok biner ( binary map ). Blok biner yang telah dihasilkan akan dijadikan sebagai masukan pembangkitan model latar ( background modelling ). Model latar bertujuan untuk memisahkan objek bergerak dengan latar yang ada pada citra masukan. Objek bergerak yang dihasilkan (lokalisasi objek) akan dikirimkan ke tahap klasifikasi objek menggunakan deep learning . Dataset yang digunakan dalam penelitian ini adalah CDNet 2014. Hasil penelitian mampu menghasilkan sistem deteksi objek Abstract — Moving object detection systems have been widely used in everyday life. Currently, research in the field of background subtraction is still being carried out to achieve maximum accuracy results. This study aims to model the background subtraction of an image using the mean value with the concept of non overlapping block. Furthermore, the background abstraction results will be used in deep learning-based moving object detection. Specifically, the input image will be divided into several blocks, then the mean value of each block will be calculated to later produce a binary block (binary map). The binary blocks that have been generated will be used as input for background modeling. The background model aims to separate moving objects from the background in the input image. The resulting moving object (object localization) will be sent to the object classification stage using deep learning. The dataset used in this study is CDNet 2014. The results of the study were able to produce a more accurate moving object detection system. Quantitative tests carried out resulted in an accuracy of above 90%.