Implementasi Deep Learning Metode Convolutional Neural Network (Cnn) Dalam Pengenalan Karakter Pelat Nomor Kendaraan Berbasis Python
Total View This Week0
Total View This Week0
Institusion
Institut Teknologi Telkom Purwokerto
Author
Alif, Hasanuddin Robbani
Subject
TA Engineering (General). Civil engineering (General)
Datestamp
2022-10-04 04:00:06
Abstract :
The development of traffic safety technology is increasing in line with the high use of motorized vehicles in Indonesia. Deep Learning as a new scientific field in the field of Machine Learning has recently developed to contribute to this development. The Convolutional Neural Network (CNN) method is called the best Deep Learning method and has high accuracy in image recognition. CNN has a similar way of working with the connection pattern of neurons or nerve cells of the human brain. In this study, there are several datasets used for the learning and texting process, including vehicle images with clearly visible license plates and a collection of characters from numbers to letters in png format. The Haur Cascade algorithm is used for location classification and separating it from images outside the license plate. Preprocessing is carried out by cropping and filtering to classify objects for later segmentation. With the CNN algorithm, training from training data is carried out. Three models were obtained with differences in the number of datasets used. The results of the character prediction test on the segmented plate obtained accuracy values on model I using 1080 data of 78.47%, model 2 used 1368 data of 83.53% and model 3 used 1980 data of 90.18% Keywords: Deep Learning Machine Learning. Licence Plate, CNN. Haar Cascade
The development of traffic safety technology is increasing in line with the high use of motorized vehicles in Indonesia. Deep Learning as a new scientific field in the field of Machine Learning has recently developed to contribute to this development. The Convolutional Neural Network (CNN) method is called the best Deep Learning method and has high accuracy in image recognition. CNN has a similar way of working with the connection pattern of neurons or nerve cells of the human brain. In this study, there are several datasets used for the learning and texting process, including vehicle images with clearly visible license plates and a collection of characters from numbers to letters in png format. The Haur Cascade algorithm is used for location classification and separating it from images outside the license plate. Preprocessing is carried out by cropping and filtering to classify objects for later segmentation. With the CNN algorithm, training from training data is carried out. Three models were obtained with differences in the number of datasets used. The results of the character prediction test on the segmented plate obtained accuracy values on model I using 1080 data of 78.47%, model 2 used 1368 data of 83.53% and model 3 used 1980 data of 90.18% Keywords: Deep Learning Machine Learning. Licence Plate, CNN. Haar Cascade
Download
book
BibTex
Latex, Jabref
cloud_download
Original Resource
url resource Institution
