In the last decade, surveillance camera technology has become widely practiced in public and private places to ensure the safety of individuals. Merely, face to limits of violation the private life of people and the inability to identify malicious persons that hid their faces, finding a new policy of surveillance video has become compulsory. The authors' work deals on the development of a suspicious person detection system using a new insect behaviour algorithm called artificial social cockroaches ASC based on a new image representation method (n-gram pixel). It has as input a set of artificial cockroaches (human images) to classify them (hide) into shelters (classes) suspicious or normal depending on a set of aggregation rules (shelter darkness, congener's attraction and security quality). Their experiments were performed on a modified MuHAVi dataset and using the validation measures (recall, precision, f-measure, entropy and accuracy), in order to show the benefit derived from using such approach compared to the result of classical algorithms (KNN and C4.5). Finally, a visualisation step was achieved to see the results in graphical form with more realism for the purpose to help policeman, security associations and justice in their investigation.
A Monocular Vision-based Specific Person Detection System for Mobile Robot Applications
