Analysis of Video Forensics System for Detection of Gun, Mask and Anomaly Using Soft Computing Techniques

Abstract

The video forensics world is a developing network of experts associated with the computerized video forensics industry. With quickly developing innovation, the video turned out to be the most significant weapon in the battle against individuals who violate the law by catching them in the act. Proof caught on video is viewed as more dependable, more exact, and more persuading than observer declaration alone. But, proof can be effortlessly tempered by utilizing programming. Video forensics examination, tells us about the accuracy of the input video. It has become a challenge for law enforcement agencies to deal with the increasing violence rate which involves the use of masks and weapons. The identification of a person becomes difficult with the use of face masks. The proposed method uses an efficient technique that is YOLO to detect guns, masks and suspicious persons from a video by extracting frames and features. It further compares the obtained frame with the available images in the dataset and generates output with bounding boxes detecting guns, masks and suspicious persons. This paper also examined the domain of video forensics and its outcomes. Experimental results show that the proposed method outperforms the existing techniques tested on different datasets. The precision for YOLO design for guns and masks is 100% and 75% respectively. The precision for customized CNN engineering for guns and face masks is 61.54% and 61.5% respectively. Execution measurements for both models have shown that the YOLO design outperformed the customized CNN with its presentation.