Computer Vision and Pattern Recognition in Forensic Science
The article discusses computer vision as a modern technology of automatic processing of graphic images, analyzes the relations between the terms «computer vision» and «machine vision». History of development of this technology is described, it occurred because of improvements in both computer technology and software. The computerization of forensic activities boils down to three areas: speeding up, simplifying, and improving the efficiency of information processing. A schema of a typical computer vision system is given, the possibility of using systems based on artificial neural networks for image analysis is considered. The current state of computer vision application systems and the possibility of its application in order to solve the problems of criminal justice are analyzed. The main areas of application of computer vision in forensic activities are identification of a person on the basis of his appearance, both during operational identification of a person and portrait examinations, photo and video examinations; quantitative assessment of objects in the image (for example, in case of calculating mass events’ participants); at preliminary and expert research of documents and their requisites; in functioning of criminal registration systems. Criteria and technical conditions for sampling signatures for creation of a training dataset for a neural network are given, the basics of developing an artificial neural network recognizing signs of signatures’ forgery is analyzed, which include three steps: creating a training dataset, adjusting weights and training priorities, testing the quality of network training.