Terrorism is still a tense issue. After blasting, or other events, the police can certainly busy specifies a list of suspects and occasionally start spreading facial images, finding people suspected of being the perpetrator.
In identifying the perpetrator's face or a fugitive, one way is to match a suspect's face with archive photographs. Problem becomes complicated if the very large photo archives and the search process is done manually. This method takes a long time.
Looking at the issue, Rahmat Widyanto, Chairman of Computer Science Graduate Program, University of Indonesia as well as Manager of Research at the Faculty of Computer Science, is now developing Face Image Retrieval System. The software was exhibited in the degree of Science and Innovation, University of Indonesia in 2007 two weeks ago. The device was developed with the peculiarities noticed Indonesian facial features.
According to Grace, the system was developed using the method Eigenface and euclidean distance. Eigenface methods used to conduct important facial feature extraction. Eigenface based on principal component analysis (PCA), which spelled out the most successful approaches to extract facial information.
This method uses the projection room with a high dimensional face image to a feature space with lower dimensions. Euclidean distance is used to measure the value of similarity between two facial images. The smaller the distance between two facial images, the higher the similarity value. Facial images in the database is extracted and stored.
"When the user enters the facial image to be identified, the system will look for similarities between the input data with the existing face images in the database," said Rahmat who earned his master's and doctorate from Tokyo Institute of Technology in the field computional intelegent it.
The system was able to find a face image that is relevant to the input image with an average accuracy rate of 87 percent of the face image database used. Facial feature extraction process is quite simple, fast, and efficient.
Rahmat said that the software can be connected with a camera which then automatically generated based on the image, the system will search if the face is in their list of banned or fugitives.
The device would be very beneficial to the world police. This facial recognition tool strategically placed at airports, special flights abroad, which is usually through desk-checking of documents. It will very useful if used at police headquarters to recognize the faces of perpetrators.
At the University of Indonesia, Rahmat is discussing with the new admission selection committee to use the tool. Use of the device is primarily to avoid the practice of jockeys. "There have been cases in the registration form photo female face, but when the revenue comes turns out to men. With the tool, it can be matched facial similarity," he said.
In laboratory and prototype phase, the software can be used. "For the database, we took a sample of 500 university students face the University of Indonesia and other images from the internet. Now living in synergy with those who need it," he said.
Without the help of illustrator
In development, the laboratory led by Rahmat also developed the Intelligent Software for Face Identification At Large with fuzzy similarity measure. The program is part of the Commodity Research, University of Indonesia, Main Courses and still in progress.
Identification of faces in the police usually by making sketches by the illustrator's face based on witness testimony. Sketches were then matched with databases in police facial images. The second stage has a big problem.
In the first stage, the accuracy of facial sketch interpretation depends strongly on the ability of illustrators and witnesses in verbal ability to express facial characteristics. In the second stage, the official photo in the database search may be susceptible to saturation because they have to compare the sketches of faces with all the photos there.
Rahmat software developed using a user friendly interface that can be used to visualize without the aid of the fugitive's face sketch illustrator. Users can instantly create and witnesses without the aid of facial sketch illustrator. They simply enter the verbal characteristics of the face that will be identified by selecting the properties of each facial feature that will be used as a component of the search. To better direct the user to select characteristics, for each property will be given the choice of an illustration in the form of sketches from the selected facial features. Problems brought about by the limited ability of witnesses in describing facial features had resolved verbally.
The important thing to note, said Rahmat, is characterized by the number of features of each face must not be filled at random. More and more options will be more confusing the witness. These features are placed on the user interface must be really able to represent the generalization of any general form of the characteristics possessed character of the face.
Face sketch is then used to search for a fugitive on the face image databases using artificial intelligence techniques is fuzzy similarity measure. This technique allows the search data with high complexity so that more accurate search results. The output of this software is a list of fugitives who face images sorted by the level of similarity. Search process will be done gradually for each facial feature that has been determined.
So far, all over the world have developed some software for facial identification, such as Faces, Faces-Composite Picture Programme, FACETTE, and Comphotofit. The device is a significant part of reconstructing some of the hair, forehead, eyebrows, eyes, cheeks, nose, mouth and jaw. Users include the Central Intelligence Agency U.S. (CIA) and U.S. Federal Bureau of Investigation (FBI).
It's just that the software has no facial features with a typical Indonesian facial features. Another disadvantage is not having a search facility based on the facial sketch facial image database.
Related research, Rahmat and his team had visited the police. For identification purposes the face, the National Police Criminal Investigation Section Identification Center published a Technical Guide No. Pol: Sketch Juknis/01/VIII/2006 about face expression. Facial characteristics contained in the book will be used as a reference matching a team that features the face of the software developed.
Rahmat said the research is expected to be completed next year. He hoped these studies are useful to society, especially for the relevant parties, such as the police so that later the fugitives a disturbing society more easily detected. (KOMPAS)
No comments:
Post a Comment