Deteksi dan Pelacakan Wajah Berdasarkan Warna Kulit Menggunakan Partikel Filter

Abstract

Face detection and tracking are one of the areas which always in progress in many applications such as security system, biometric and so on. However, face detection and tracking can become a complex problem when it is done in real time condition. Many problems are appeared such as dependency the detection result to lightness and diversity of background colour. In this article, we presented a new method to overcome the above problems by using the HSL colour filter (Hue, Saturation, Luminance) combined with the particle filter algorithm. The HSL colour filter detected face based on skin colour with adjusting the threshold of each H, S and L colour. The next step is face tracking using particle filter. It initialised by generating the particle filter randomly around the detected face. The weight of each particle was calculated based on a distance of each particle to face position and resampled to obtain a new particles distribution. The particle with high weight would be selected many times, and particle with small weight would be disappeared. The last process in particle filter is resampling process which was done by averaging the particle's position. The experimental results showed the average error of face tracking using 100, 200 and 500 particles on X and Y axis were (1.25; 1.02), (0.86; 0.81) and (0.59; 0.57) respectively.