Crime Detection System with Machine Learning using OpenCV, YOLO and CNN
N. V. Rajasekhar Reddy, P Priya, Shwetank Aryan, P Ajay · 2023
The number of crimes are rising daily in the current environment. The government has put CCTV cameras in several public locations to help identify suspects. It can be difficult to recognise offenders on the footage, though. This system has created a procedure to speed up this process and lower the rising crime rate. In comparison to conventional approaches, this technology provides superior solutions and automates the suspect detection procedure. With the aid of a distributed CCTV system, This system that records the faces, gestures, and emotions of the targeted people (criminals). Together with a time and location stamp, will keep a database containing this information. The generated database will be used to identify suspects using CCTV footage of crimes that was recorded by a number of systems situated along routes and close to the crime site. This study examines several techniques including face identification using the OpenFace algorithm, object detection using the YOLOv3 algorithm, and hand motion recognition using the CNN algorithm for locating suspects in CCTV footage and turning that data into information that may be utilised to analyse certain crime scenes in more detail. The findings of our investigation show how successful our method is at identifying suspects. For identifying the expressions of anger, happiness, and neutrality, the OpenFace algorithm obtains accuracy of 55%, 60%, and 80%, respectively. The YOLOv3 algorithm successfully and accurately identifies firearms. The CNN algorithm recognises hand movements precisely, making it possible to identify certain motions or indications. Our research shows that the distributed CCTV system that is presented, when combined with algorithms, offers greater options for suspect detection and identification. It improves the precision of suspect identification and lowers the incidence of false alarms by automating the process and using contextual information. Comprehensive analysis of CCTV video is made possible by the integration of facial recognition, object identification, and hand gesture recognition, enabling a more in-depth comprehension of crime scenes. The system sends a report to the police in the form of a database whenever a crime is discovered. The result is sent to the police through Email in this suggested approach. Simple Mail Transfer Protocol, the industry-standard protocol for internet communication and the transmission of electronic mail, is used to transmit mail. In conclusion, this study makes a significant addition to the fields of crime detection and surveillance technologies. Suspect identification is made more efficient and effective by using a distributed CCTV system using algorithms. This system provides law enforcement organisations a useful tool to fight crime and improve public safety by utilising computer vision. The ability to analyse and understand CCTV footage for criminal investigations might be revolutionised by further research and development in this field that results in ever more sophisticated systems.