An Improved Crime Scene Detection System Based on Convolutional Neural Networks and Video Surveillance

Since criminals may engage in a wide variety of crime scenes in public spaces, the time immediately before and after these events must be monitored. As a rule, there will be cameras set up for monitoring purposes. Videos or still photographs captured by these cameras would reveal any criminal behavior to authorities after the fact. Consequently, many people many look for signs of fraud. However, there are currently no safeguards to prevent the theft of high-value items. Pre-incident warning alarm systems based on ML and DL may now be customized to track crime behavior. Human behavior analysis and identifying out-of-the-ordinary events would provide the basis for predicting the occurrence of crime scenes The necessity for a highly accurate, precise, low-false-positive, and low-false-negative prediction system has led to the development of several ML and DL-based techniques; however, a hybrid or upgraded ML- or DL-based system may better answer this demand. A better criminal activity detection system based on an enhanced convolutional neural network (E-CNN) is presented in this proposed study. It was stated that the experiment was successful. The SPSS software is used to examine the data. The findings revealed that crime activity detections had an average accuracy of 97.050%, precision of 96.743 %, the false-positive rate of 2.957%, and false-negative rate of 2.927%. The convolutional neural network (CNN) method was also compared to this outcome. The findings of this study may be used to better detect and prevent potentially dangerous situations by enhancing the security warning system for potential criminal activities.