Implementasi Sistem Monitoring Kehadiran Otomatis Berbasis IoT Menggunakan Sensor Sidik Jari dan Kmaera

Abstract

PT MAE Abadi Evolusi mengalami inefisiensi pada pencatatan kehadiran karyawan: pemindai sidik jari yang digunakan sering gagal baca, sedangkan prosedur cadangan berupa pengiriman foto selfie melalui WhatsApp lambat diverifikasi dan menyulitkan rekapitulasi data. Penelitian ini merancang dan mengimplementasikan sistem absensi multifaktor berbasis Raspberry Pi 4 Model B yang menggabungkan sensor sidik jari optik DFRobot SEN0188 dan pengenalan wajah melalui pustaka face_recognition (embedding 128 dimensi, tolerance 0,5, margin 0,08), dilengkapi dashboard web Flask dan basis data SQLite lokal. Citra wajah diambil paralel dengan pemindaian sidik jari dan hanya diproses apabila sidik jari gagal dikenali. Pengujian mencakup uji fungsionalitas dengan tujuh partisipan dan uji ketahanan pose kepala dengan lima partisipan (255 percobaan). Modul sidik jari mencapai tingkat keberhasilan 88,6% dengan latency ratarata 793 milidetik, sedangkan pengenalan wajah 80,0% dengan 240 milidetik; nilai 240 milidetik merupakan biaya pemrosesan tambahan karena citra wajah telah diperoleh secara paralel, bukan waktu akuisisi hingga pengenalan dari kondisi awal. Skema multi-faktor mencapai keberhasilan absensi 100% karena seluruh tujuh partisipan tetap dapat absen meski salah satu modul gagal. Uji ketahanan pose kepala menghasilkan keberhasilan agregat 55,3% tanpa satu pun false accept. Sistem berhasil diimplementasikan dan berjalan otomatis; pendekatan multifaktor terbukti meningkatkan keberlangsungan autentikasi dibandingkan modalitas tunggal.

PT MAE Abadi Evolusi experienced inefficiencies in employee attendance recording: the fingerprint scanner in use frequently failed to read, while the backup procedure of sending selfie photos through WhatsApp was slow to verify and hard to consolidate. This study designed and implemented a multi-factor attendance system on a Raspberry Pi 4 Model B, combining a DFRobot SEN0188 optical fingerprint sensor with face recognition via the face_recognition library (128-dimensional embeddings, tolerance 0.5, margin 0.08), with a Flask web dashboard and a local SQLite database. The facial image is captured in parallel with the fingerprint scan and processed only when the fingerprint fails. Testing covered functionality (seven participants) and head pose robustness (five participants, 255 trials). The fingerprint module achieved 88.6% success at an average latency of 793 milliseconds, and face recognition 80.0% at 240 milliseconds; the latter is the incremental processing cost, because the image had already been acquired in parallel, not the acquisition-to-recognition time from an idle state. The multi-factor scheme achieved 100% attendance success, as all seven participants could still record attendance when one module failed. Head pose robustness testing produced an aggregate 55.3% with no false accept. The system was implemented successfully and runs automatically; the multifactor approach improved authentication continuity over a single modality.