Abstract
Palm print is one of the organs of the human body that can be used as identification. Palm print has a unique characteristic, hard to forge and more reliable than finger print, and identify a person using the palm print as any other biometrics. In this research we proposed palm print identification using K-Nearest Neighbor (KNN). The identification consisted of data acquisition, intensity normalization result of image pre-processing segmentation, feature extraction and classification. The proposed method is trained and evaluated using 900 images of 10 individuals with a 70:30 percentage containing images sizes 32 × 64, 64 × 128 and 128 × 256. Pre-processing using canny edge detection prosed Also to reduce the noise in the images. The features Including contrast, energy, entropy and homogeneity on 900 extracted using Gray-level co-occurrence matrix (GLCM). Finally the result of pre-processing is classified using KNN with k parameter that compute the distance of neighbors (similar data) between 3, 5, and 7. The evaluation resulted the best number of k for this dataset is 7 the which is the highest number. Image size 32 × 64 give the fastest time in the extraction and classification process, the which is 47.04 seconds. The accuracy of KNN classifier is evaluated using k-fold cross validation and archived 98% of accuracy as higher accuracy with k = 7. In the future the classification need to re-evaluated using higher number of k and more variety of image data sets. Image size 32 × 64 give the fastest time in the extraction and classification process, the which is 47.04 seconds. The accuracy of KNN classifier is evaluated using k-fold cross validation and archived 98% of accuracy as higher accuracy with k = 7. In the future the classification need to re-evaluated using higher number of k and more variety of image data sets. Image size 32 × 64 give the fastest time in the extraction and classification process, the which is 47.04 seconds. The accuracy of KNN classifier is evaluated using k-fold cross validation and archived 98% of accuracy as higher accuracy with k = 7. In the future the classification need to re-evaluated using higher number of k and more variety of image data sets.
Concepts :
Access to Document
SDGs
Citations by Year
| Year | Count |
|---|---|
| 2020 | 4 |