Tugas Akhir ini membahas mengenai cara mendeteksi penyusup melalui video yang direkam menggunakan kamera infra merah. Beberapa penelitian telah dilakukan agar kamera bisa mendeteksi penyusup, salah satunya adalah yang dilakukan oleh Picolov. Menurut Picolov, penyusup ditentukan berdasarkan ukuran objek yang terdeteksi. Asalkan objek tersebut berukuran besar maka objek tersebut penyusup. Pendeteksian penyusup yang lebih ideal mampu membedakan objek yang terdeteksi merupakan orang atau bukan, sehingga bisa mengurangi false alarm. Pada Tugas Akhir, ini diusulkan cara mendeteksi penyusup berdasarkan face detection dan human detection. Metoda face detection yang digunakan adalah metoda Viola-Jones. Sedangkan untuk human detection menggunakan Histogram of Oriented Gradient (HOG). Pendeteksian dilakukan secara tidak real time dan offline, bersumber pada rekaman video. Pendeteksian terdiri dari 3 tahap. Pertama pembuatan background image. Kedua, ekstraksi objek. Ketiga, melakukan face detection dan human detection untuk mendeteksi penyusup. Uji coba pendeteksian dilakukan terhadap video yang direkam menggunakan CCTV infra merah di koridor lantai 2 Gedung Jurusan Teknik Komputer Polban. Hasil uji coba menunjukkan pendeteksian penyusup dengan akurasi 61,94%, dan kecepatan pendeteksian rata-rata 1,38 Frame Per Second (FPS) untuK video dengan resolusi 1280x720 pixel. Kata kunci: pendeteksian penyusup, face detection, Viola-Jones, human detection, Histogram of Oriented Gradient (HOG).

This Final Project discussed about how to detect intruder through video which recorded with infrared camera. Some research about intruder detection had been done several times, one of them was the Picolov’s research. According to Picolov, intruder was detected based on its size. As long as the detected object is big enough, then that object was intruder. Intruder detection called ideal if it can determine between human or non-human, so it can reduce false alarm. This Final Project proposed a method of intruder detection based on face detection and human detection. The face detection method used was the Viola-Jones method, and for the human detection method used was Histogram of Oriented Gradient (HOG). The detection is not real time and offline, the source was recorded video. Intruder detection was divided into three steps. First step was the reconstruction of background image. The second step was object extraction. And the third was intruder detection based on face detection and human detection. The experiment was done toward the video tape which recorded using infrared CCTV at the corridor of 2nd Floor Jurusan Teknik Informatika Building Polban. Result of the experiment shows that the intruder detection has an accuracy of 61,94% and detection speed of 1,38 frame per second (FPS) for video with 1280x720 resolution. Keywords: intruder detection, face detection, Viola-Jones, human detection, Histogram of Oriented Gradient (HOG).

http://digilib.polban.ac.id/gdl.php?mod=browse&op=read&id=jbptppolban-gdl-fajarkhari-5722