An integrated technique for face sketch recognition using DCNN

Abstract

Face Recognition (FR) is considered as one of the chief use in the investigation of criminals. In the majority of the cases, information about the criminal is not available. In such situations, sketch artist draw the sketch of the guess with the oral explanation provided by the eyewitness. These sketches can then be matched manually against mug shot photos. This process is time-consuming. Hence there require a method that efficiently goes with composite sketches to the gallery of mug shot databases. Thus the proposed system uses a scheme for matching composite sketch and photo images, photo image features are extracted and fused to train the system. Composite Sketch feature is matched with face photo images. Feature extraction (FE) is done using Multi-Scale Local Binary Patterns (MLBP) Tchebichef Moments and Multiscale Circular Weber Local Descriptor (MCWLD), Principal Component Analysis (PCA) is used for fusion of extracted features, DCNN used as a classifier to recognize the face. The experiments are conducted using PRIP-HDC dataset and the proposed system gives good accuracy in face recognition.