SIGNATURE-BASED DOOR LOCK SYSTEM USING KNN ALGORITHM

Abstract

: The Signature verification plays a crucial role in authentication of door lock system, which provides a secured door lock system replacing the use of mechanical lock, scanner card, memorizing password and OTP. It also instills more secured feeling in the user’s mind. In this paper, the application of the K-nearest neighbors (KNN) algorithm in signature verification has been explored while using canny edge features extracted from signature images. The project has been conducted by preprocessing a dataset of ten individuals’ signature images while extracting canny edge features to represent the unique characteristics of each signature. Subsequently, trained the KNN algorithm using a subset of the dataset, memorizing the feature-label associations. During the prediction phase, utilized the trained KNN model to classify new signature images based on the individual canny edge features. In this work, grey image has extracted from the original image, from that grey image will be converted into edge image, from that Histogram of Oriented Gradients (HOG) feature descriptor is used to extract features from the edge-detected image, this will do for both reference image and verification image and mean feature vector will also calculate and then both images will be compared then the authentication process will be done. By comparing the ‘K’ values of k=1 and k=3, Our findings highlight the significance of ‘k’ in influencing the accuracy and reliability of signature verification which is crucial to increase the accuracy of door lock security system and there by optimal selection of ‘k’ for different applications.