Soft Lens Detection in Iris Image using Lens Boundary Analysis and Pattern Recognition Approach

Recent studies have demonstrated that the soft lens wearing during iris recognition has indicated the increase of false reject rate. It denies the strong belief that the soft lens wearing will cause no performance degradation. Therefore, it is a necessity for an iris recognition system to be able to detect the presence of soft lens prior to iris recognition. As a first step towards soft lens detection, this study proposed a method for segmenting the soft lens boundary in iris images. However, segmenting the soft lens boundary is a very challenging task due to its marginal contrast. Besides, the flash lighting effect during the iris image enrolment has caused the image to suffer from inconsistent illumination. In addition, the visibility condition of the soft lens boundary may be discerned as a bright or dark ridge as a result of the flash lighting. Three image enhancement techniques were therefore proposed in order to enhance the contrast of the soft lens boundary and to provide an even distribution of intensities across the image. A method called summed-histogram has been incorporated as a solution to classify the visibility condition of the soft lens boundary automatically. The visibility condition of the ridge is used to determine the directional directive magnitude by the ridge detection algorithm. The proposed method was evaluated with Notre Dame Contact Lens Detection 2013 database. Results showed that the proposed method has successfully segment the soft lens boundary with an accuracy of over 92%.

Implementasi Webcam sebagai Pendeteksi Wajah pada Sistem Keamanan Perumahan menggunakan Image Processing

Pencurian yang sangat meresahkan masyarakat seringkali terjadi pada suatu ruangan atau lingkungan seperti gedung, kantor, lorong bahkan tempat ibadah juga menjadi sasaran para pencuri. Upaya yang dilakukan DKM (Dewan Kemakmuran Masjid) yaitu memberikan himbauan supaya tetap menjaga barang pentingnya masing-masing. Masjid seharusnya menjadi tempat yang aman dan nyaman untuk dikunjungi. Oleh karena itu kami memiliki ide yang bertujuan untuk mengantisipasi pencurian di masjid atau tempattempat yang rawan pencurian. Penelitian ini merancangbangun sistem pengenalan wajah sebagai solusi untuk mengurangi tingkat pencurian. Sistem ini dilengkapi dengan perangkat keras Raspberry Pi 3 model B dan webcam A4Tech. Perangkat lunak database yang dapat menyimpan data pengguna. Tujuan penelitian untuk membandingkan 2 metode yang terbaik dalam pengenalan wajah yaitu metode LBPH (Local Binary Pattern Histogram) dan metode Eigenface. Penelitian dilakukan pada siang hari untuk mengambil citra wajah yang berbeda. Penelitian dilakukan dengan 3 kondisi yaitu siang hari luar ruangan, siang hari dalam ruangan dan malam hari dalam ruangan. Parameter yang digunakan untuk melihat hasil dari pengenalan wajah yaitu Akurasi, FAR (False Accept Rate) dan FRR (False Reject Rate). Hasil pengujian 2 metode tersebut yang memiliki tingkat rata-rata Akurasi tertinggi dan tingkat rata-rata FAR dan FRR terendah adalah metode Eigenface. Kesimpulan dari hasil penelitian yaitu pencahayaan mempengaruhi pengenalan wajah dalam 2 metode tersebut.

Contact Lens Classification by Using Segmented Lens Boundary Features

Recent studies have shown that the wearing of soft lens may lead to performance degradation with the increase of false reject rate. However, detecting the presence of soft lens is a non-trivial task as its texture that almost indiscernible. In this work, we proposed a classification method to identify the existence of soft lens in iris image. Our proposed method starts with segmenting the lens boundary on top of the sclera region. Then, the segmented boundary is used as features and extracted by local descriptors. These features are then trained and classified using Support Vector Machines. This method was tested on Notre Dame Cosmetic Contact Lens 2013 database. Experiment showed that the proposed method performed better than state of the art methods.

AN ASYMMETRIC ADAPTIVE CLASSIFICATION METHOD

There should be different requirements for False Reject rate and False Accept rate in classification applications, and classifier learning should use an asymmetric factor to balance between False Reject rate and False Accept rate. A novel AdaBoost algorithm was developed with the asymmetric weight. Moreover we provide the theoretical analysis of its performance and derive the upper bound of the classification error.