Februari 26, 2011
How to recognize fake autographs
Signature forger sniff the "Software"
Generally, identifying signatures done manually, which is suspected of comparing signatures with the original. With the software, counterfeiting can be known quickly and precisely.
Achmad Hidayatno, professor of Electrical Engineering of Diponegoro University, Semarang, infuriated when his signature was forged student card to take care of the study plan (KRS). Improper behavior of students was exposed when officers see the clumsiness administration Achmad signature. To sniff out the authenticity of his signature, the officer was forced to match the original signature Achmad with a forged. As a result, false signatures are similar to the original, but there are scratches that looked less graceful pen. Finally the officer asked for confirmation to Ahmad to ascertain whether the right ever signed in KRS belonging to one of his students. Evidently Achmad was never given such endorsement signature on the KRS students.
To anticipate the same event, lecturers who have an interest in digital image processing is then to design software (software) to identify specific signatures. Identification is an important process to identify the characteristic signature, which is certainly different from one person to another.
Development of the basic method of identification in this digital signature using the natural characteristics of the human hand as a base scratches which became known as biometrics. Comparison of the accuracy of biometric technology, which refers to the comparison between the fault identification process with the accuracy of the identification process is one in 50 (150).
However, to get the output of biometric technology, it takes a data base of the original signature first. "The more the sample signature for the purposes of data base, output is expected from the application of biometric technologies will be more accurate," said Achmad.
Because, the person's signature is also influenced by psychological conditions. That is, even if only one person, often times the resulting signature persen.Sistem not correspond 100 signatures made by Achmad identification using neural network model of tin-back propagation (back-propagation neural network) built with Delphi programming language.
"This application serves to change the data input in the form of sample signatures into digital form digit (binary). Then the data is processed Aruan neural network system to determine the authenticity of a person's signature," said Achmad.
Processing Data
At first, the sample signature is used as a reference was added in a sheet of paper that the format has been adjusted. The format was such a box as a limit when the appended signature. All samples etched using a pen with black ink. Signature scan is then performed with a scanner tool to convert the data into digital data.
From the results of the scan, it can be pre-processing (preprocessing) and extraction. In the pre-processing carried out changing the image resolution, image contrast, and image edge detection. While in the process of extraction is done by segmenting the image in the form of rows and columns.
The goal is to obtain significant information from the image feature signatures, and to obtain data values that will be input to the neural network.
Examples of the extraction process is the signature images that were previously located in a box subdivided into OGA rows and three columns, so that a nine smaller squares. Each box is enlarged pixels to determine the value of low intensity in it.
In this case, low-intensity identical to the dark color, eg black. If the box is valued at a low intensity (black), then the box would be worth 1. However if the value of high intensity (white), then the box would be worth 0. "Then the data in the form of the numbers are used as inputs in tin propagation neural network and forth," said Achmad.
Neural networks tin, clear Achmad, is a way of solving the problem by adopting work procedures of the human brain consists of millions of nerve cells. The brain is a complex tool, non-linear, and has a parallel process that can process the input signal into an output that can be identified for further processing.
Thus, neural networks in general is a machine designed as a model, as the brain do its work that can be either hardware or software.
In designing a neural network, in addition to considering the structure of the relationship between the input data with output data, is necessary to determine also the way or method of learning. How this neural network learning is to provide the data called training data, consisting of pairs of input and
the desired output. After the training process is complete, then given a neural network input values and will produce the desired output.
To support this research, Achmad collect 150 signatures as a data base obtained from 10 respondents. Each respondent to sign as many as 10. As a result, this application system has a success rate of 95 percent accuracy.
Generally, identifying signatures done manually, which is suspected of comparing signatures with the original. With the software, counterfeiting can be known quickly and precisely.
Achmad Hidayatno, professor of Electrical Engineering of Diponegoro University, Semarang, infuriated when his signature was forged student card to take care of the study plan (KRS). Improper behavior of students was exposed when officers see the clumsiness administration Achmad signature. To sniff out the authenticity of his signature, the officer was forced to match the original signature Achmad with a forged. As a result, false signatures are similar to the original, but there are scratches that looked less graceful pen. Finally the officer asked for confirmation to Ahmad to ascertain whether the right ever signed in KRS belonging to one of his students. Evidently Achmad was never given such endorsement signature on the KRS students.
To anticipate the same event, lecturers who have an interest in digital image processing is then to design software (software) to identify specific signatures. Identification is an important process to identify the characteristic signature, which is certainly different from one person to another.
Development of the basic method of identification in this digital signature using the natural characteristics of the human hand as a base scratches which became known as biometrics. Comparison of the accuracy of biometric technology, which refers to the comparison between the fault identification process with the accuracy of the identification process is one in 50 (150).
However, to get the output of biometric technology, it takes a data base of the original signature first. "The more the sample signature for the purposes of data base, output is expected from the application of biometric technologies will be more accurate," said Achmad.
Because, the person's signature is also influenced by psychological conditions. That is, even if only one person, often times the resulting signature persen.Sistem not correspond 100 signatures made by Achmad identification using neural network model of tin-back propagation (back-propagation neural network) built with Delphi programming language.
"This application serves to change the data input in the form of sample signatures into digital form digit (binary). Then the data is processed Aruan neural network system to determine the authenticity of a person's signature," said Achmad.
Processing Data
At first, the sample signature is used as a reference was added in a sheet of paper that the format has been adjusted. The format was such a box as a limit when the appended signature. All samples etched using a pen with black ink. Signature scan is then performed with a scanner tool to convert the data into digital data.
From the results of the scan, it can be pre-processing (preprocessing) and extraction. In the pre-processing carried out changing the image resolution, image contrast, and image edge detection. While in the process of extraction is done by segmenting the image in the form of rows and columns.
The goal is to obtain significant information from the image feature signatures, and to obtain data values that will be input to the neural network.
Examples of the extraction process is the signature images that were previously located in a box subdivided into OGA rows and three columns, so that a nine smaller squares. Each box is enlarged pixels to determine the value of low intensity in it.
In this case, low-intensity identical to the dark color, eg black. If the box is valued at a low intensity (black), then the box would be worth 1. However if the value of high intensity (white), then the box would be worth 0. "Then the data in the form of the numbers are used as inputs in tin propagation neural network and forth," said Achmad.
Neural networks tin, clear Achmad, is a way of solving the problem by adopting work procedures of the human brain consists of millions of nerve cells. The brain is a complex tool, non-linear, and has a parallel process that can process the input signal into an output that can be identified for further processing.
Thus, neural networks in general is a machine designed as a model, as the brain do its work that can be either hardware or software.
In designing a neural network, in addition to considering the structure of the relationship between the input data with output data, is necessary to determine also the way or method of learning. How this neural network learning is to provide the data called training data, consisting of pairs of input and
the desired output. After the training process is complete, then given a neural network input values and will produce the desired output.
To support this research, Achmad collect 150 signatures as a data base obtained from 10 respondents. Each respondent to sign as many as 10. As a result, this application system has a success rate of 95 percent accuracy.
Langganan:
Posting Komentar (Atom)
Tidak ada komentar:
Posting Komentar