Perancangan Aplikasi Pengenalan Gambar Objek Dan Perangkat Pemantauan Menggunakan Metode Convolutional Neural Network

Abstract

Abstrak Sistem monitoring pada perkembangan teknologi sekarang ini sangat dibutuhkan untuk keamanan suatu ruangan. Salah satu sistem monitoring adalah kamera CCTV. Kamera CCTV banyak digunakan pada perkantoran, instansi militer, rumah sakit, bank dan lain – lain. Kamera CCTV merekam semua kejadian yang ada pada suatu ruangan selama 24 jam. Namun hal itu akan mengakibatkan pemborosan memori penyimpanan data. Maka dari itu diimplementasikan metode pengolahan citra yaitu deteksi objek. Penulis menggunakan IP Camera sebagai kamera CCTV guna memberikan gambar hasil deteksi. Dalam pelaksanaannya penulis menggunakan metode Convolutional Neural Network (CNN) untuk mengenali gambar apabila adanya objek yang terdeteksi , lalu hasil deteksi akan dikirim ke server dan dapat diakses melalui aplikasi pada mobile device. Pengujian menggunakan metode Convolutional Neural Network ini didapatkan analisa dengan parameter True Positive Rate (TPR), False Positive Rate (FPR), Percentage Correct Classification (PCC), dan fungsionalitas aplikasi Android pada mobile device. Fungsionalitas disini mencocokkan kebenaran hasil prediksi pengenalan gambar dan notifikasi aplikasi Android pada mobile device. Dari hasil training 2000 gambar dataset (1000 gambar human, 1000 gambar non-human) dalam pembuatan model didapatkan nilai True Positive Rate = 0.9, False Positive Rate = 0.16, dan Percentage Correct Classification = 86.6% Kata kunci : identifikasi, image processing, ip camera, mobile device, server. Abstract The monitoring system in the era of technological development is now needed for security in a room. One of the monitoring systems is CCTV camera. CCTV cameras are widely used in offices, military agencies, hospitals, banks and others. CCTV cameras record all events in a room for 24 hours. But that will cause a waste of data storage memory. Therefore, we want to implement an image processing method that is object detection. The author uses the IP camera as a CCTV camera. In the implementation the writer uses the Convolutional Neural Network (CNN) method as a method used to recognise the image if there’s a detected object. Then the detection results will be sent to the server and can be accessed through applications on the mobile device. This research is using the Convolutional Neural Network method were obtained by analyzing the parameters of True Positive Rate (TPR), False Positive Rate (FPR), Percentage Correct Classification (PCC), and the functionality of Android applications on mobile devices. The functionality here matches the truth of the prediction results of image recognition and notification of Android applications on mobile devices. From the training results of 2000 dataset images (1000 human images, 1000 non-human images) in the making of the model obtained True Positive Rate = 0.9, False Positive Rate = 0.16, and Percentage Correct Classification = 86.6% ISSN : 2355-9365 e-Proceeding of Engineering : Vol.7, No.1 April 2020 | Page 120 Keywords : recognition, image processing, ip camera, mobile device, server.