Pushpa Gopal Ambhore, Lokesh Bijole
Abstract: The evolution of biometric technology has provided criminal investigators additional tools to determine the criminals identity. Besides DNA and circumstantial evidence; if a dormant fingerprint is found at an investigative sight or a surveillance camera captures an image of a suspects face; then these cues may be used to determine the identity of culprits; using automated biometric identification. However; most of the times crimes occur when no such information is available; but in its place an eye-witness of the crime is on hand. In such situation a forensic artist is often used to work with the witness or the victim in order to draw a sketch that describes the facial appearance of the culprit according to the verbal description; such sketch drawn depending on the description given by an eye witness is called forensic sketch. Using feature based approach how to match a forensic sketch to a gallery of mugshot images described here. To match forensic sketches against mugshot images a robust framework called local feature-based discriminant analysis (LFDA) is used. Forensic sketches can be of poor quality; so to improve the quality of images and the identification performance a pre -processing technique is used. In this paper experiments are performed using 45 forensic sketches for matching against a gallery of 150 photo images. Sketch and photo images are used for extracting feature where we use LFDA framework which uses feature descriptor such as scale invariant feature transform (SIFT) and multiscale local binary pattern (MLBP) method. The experimental results demonstrate that the proposed algorithm with pre-processing approach gives enhanced identification accuracy.
Keywords: Mugshot, forensic sketch, Local feature-based discriminant analysis, Feature-based approach, Texture descriptors, Feature descriptors