Presentation is loading. Please wait.

Presentation is loading. Please wait.

Designing and Making Application To Recognize Signatures Using Backpropagation Ronny Harianto 26406037.

Similar presentations


Presentation on theme: "Designing and Making Application To Recognize Signatures Using Backpropagation Ronny Harianto 26406037."— Presentation transcript:

1 Designing and Making Application To Recognize Signatures Using Backpropagation Ronny Harianto

2 Background Signatures can be interpreted as one form of identification. Every people have different signatures, because their intensity level, writing style, and proportion. Because they have unique signatures, then I make an application to recognize the signatures.

3 Theory Thresholding  Thresholding is a process to change the value pixel of the image to 0 and 255(bi-level image). Noise Reduction  Noise Reduction is use to clean the image from noise which has form of point. Segmentation  Segmentation is use to retrieve the required object from the image.

4 Theory Width Normalization  Width nomalization is use to uniform the size of image. Thinning  Thinning is used to retrieve a structure from image. Region  Region is used to convert an image to a value that can be read by backpropagation  Image is divide into 12 regions

5 Theory Backpropagation  Backpropagation is the one of the learning algorithm in artificial neural network.  Learning process is done by adjusting the weight of the artificial neural networks with backward direction based on the error value in the learning process.

6 Design System And Application

7 Process Image Form

8 Training Form

9 Recognize Form

10 Result for signatures drilled With 10 hidden layer on first layer (90%) NamaJumlah sampel yang dikenali Jumlah sampel yang tidak dikenali Persentase Keberhasilan David45590% Ivan47394% Merlina45590% Nyoto381276% Riky44688% Ronny46492% Yohanes500100%

11 Result for signatures drilled With 20 hidden layer on first layer ( %) NamaJumlah sampel yang dikenali Jumlah sampel yang tidak dikenali Persentase Keberhasilan David47394% Ivan500100% Merlina49198% Nyoto45590% Riky45590% Ronny46492% Yohanes500100%

12 Result for signatures wasn’t drilled With 10 hidden layer on first layer ( %) NamaJumlah sampel yang dikenali Jumlah sampel yang tidak dikenali Persentase Keberhasilan David50100% Ivan50100% Merlina50100% Nyoto4180% Riky50100% Ronny4180% Yohanes50100%

13 Result for signatures wasn’t drilled With 20 hidden layer on first layer ( %) NamaJumlah sampel yang dikenali Jumlah sampel yang tidak dikenali Persentase Keberhasilan David50100% Ivan50100% Merlina50100% Nyoto4180% Riky50100% Ronny4180% Yohanes50100%


Download ppt "Designing and Making Application To Recognize Signatures Using Backpropagation Ronny Harianto 26406037."

Similar presentations


Ads by Google