AN ACCURATE FINGERPRINT RECOGNITION ALGORITHM BASED ON HISTOGRAM ORIENTED GRADIENT (HOG) FEATURE EXTRACTOR

Biometrics is a research in which the identity of an individual is based on a vector that contains features derived from a behavioural or physical characteristics. Fingerprint recognition for personal identification and authentication is one of the most well-known and published biometric. Fingerprint recognition has been studied for many years now with technological advancement and for safety in various civil, defence and commercial applications. Studies on different aspects and properties of fingerprints have been conducted since the second millennium when fingerprints were used for signature purposes. This paper proposes an accurate fingerprint model based on the oriented gradients, which can accurately identify individuals. The model proposed uses the HOG descriptor to present the fingerprint. The HOG descriptor counts the number of times in a localised fingerprint area the gradient orientation occurring. It uses a histogram of gradient intensity to show the image shape. This technique is durable when shadow and light are changed. The HOG algorithm descriptor implementation method is provided as follows. First, the fingerprint is broken up into the smallest possible image regions. These areas are termed cells, A histogram of gradient orientation is calculated for each of these cells. Each cell is segregated and separated by its gradient orientation into the appropriate angular bins. Each cell's weighted gradient contributes to its respective angular bin. The adjacent gradient-orientated cell is grouped and these spatial areas are called blocks. These clusters into blocks form the basis to normalise histograms. The normalised group is the histogram block, representing the descriptor in turn. The findings suggest that the descriptor of the HOG has substantial features that achieves an average accuracy rate of 100% for the recognition of fingerprint.