A Heterogeneous Face Recognition Approach for Matching Composite Sketch with Age Variation Digital Images

摘要

Law enforcement agencies use facial composite sketches as a forensic tool to identify the suspects involved in criminal activities. Criminal investigators use either hand drawn or software generated sketch for investigation. Matching facial composite sketch with its digital images across different age variations is an important challenge of heterogeneous face recognition. Current research works are mainly focusing on the automation of matching facial composite sketch to digital images. A framework for mapping of facial composite sketch to digital images across different age variations is proposed in this work. In the proposed framework, composite sketch is fed as input to Convolutional Neural Networks (CNN), face encoding techniques are applied to extract the important features from the face and best matching index is computed. Then bounding box is drawn for recognizing digital images of different age variations. To study the effects of age variations, experiments are conducted on the composite sketch with age variation dataset (CSA) which consist of composite sketches of 150 subjects with digital images of different age variations. The system performance is appraised through exhaustive testing and results analysis. Experimental result proves that proposed framework provides promising results of 86.6% compared with other existing sketch based algorithms.