Comparison of Matching Score Based Fusion in Latent Fingerprint Recognition

In this study, we present a comparison study towards improving latent fingerprint image recognition accuracy using fusion techniques on various minutia extraction templates. Image enhancement techniques applied to latent fingerprint images remove noise and yields a novel minutiae template for these images. It is shown that combining fingerprint matching scores obtained for different templates using fusion methods improves latent image recognition performance considerably. Fusion of minutiaes templates obtained as a result of giving the images enriched with the STFT method as input to the FingerNet method, the template obtained as the input of the raw fingerprint images to the FingerNet, and the minutiae obtained with the autoencoder approach resulted in the best performance.