Abstract

Pengenalan suatu objek secara otomatis adalah suatu pekerjaan yang sangat penting seperti halnya untuk mengidentifikasi sebuah objek tertentu. Jika hal ini dilakukan oleh manusia maka akan sulit untuk mendapatkan hasil yang baik dengan konsisten, oleh sebab itu digunakan komputer. Komputer dapat mengenali objek selayaknya kemampuan manusia dalam mengenali objek, dengan cara mengamati gambar yang diperoleh dari kamera, dan menerapkan metode pengenalan pada gambar tersebut. Pada penelitian ini metode pengenalan objek akan dikembangkan dengan menggunakan kamera fisheye yang memiliki luas tangkap empat kali kamera konvensional. Metode pengenalan objek yang digunakan yaitu deep learning dengan arsitektur CNN (Convolution Neural Network). CNN memiliki kemampuan untuk mengenali objek dalam gambar. Model CNN yang digunakan terdiri dari 1 layer, 2 layer, 3 layer, dan 7 layer. Sedangkan untuk melatih dan memvalidasi model tersebut digunakan 900 gambar dataset. Hasil pengujian pada penelitian ini menunjukan bahwa pada 7 layer CNN menghasilkan nilai presisi, recall dan akurasi tertinggi dengan komposisi nilai presisi 98,56%, recall 98,5% dan akurasi 98,59%. Nilai tersebut menunjukan bahwa hasil klasifikasi terhadap ketiga klasifikasi objek gambar manusia pada gambar fisheye dapat dilakukan dengan sangat baik. Abstract Object recognition is an important method of identifying object in an image. If this method is carried out by the humans, it is difficult to work continuously, therefore, it can be done by a computer. The computer can recognize an object on an image taken from a camera as if a human if it is trained whit a certain dataset of be object. In this research, a method a recognizing object is developed by using images captured from a fisheye camera. This camera provides better field of view four times then confessional camera. The object recognition method relies on convolutional neural network architecture that has capability to recognize object on an image. The CNN model is developed in four types of models. The first model consists of 1 layer, while the second model consist of 2 layers. The thread model consists of 3 layers, while the last consist of 7 layers. The number of datasets used for training and testing each model is 900 images. The experiment results showed that the four model which consist of 7 layers provide the best result. This is confirmed by the number of precisions, recall, and accuracy, which reach 98,56%, 98,5%, and 98,59% respectively. This result means that can classify three different objects of human on fisheye images well.

Full Text
Paper version not known

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call

Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.