Detection of Bullet Impact using Mask RCCN in Silhouettes

Bullet detection in the silhouette has a considerable influence in institutions with a greater shooting range. Particularly when there is a controlled permission to carry a handgun and the evaluation is a manual method, there may be irregularities in the shooter's capacity to analyse outcomes in the process of issuing a license. This research work presents a method for automatically detecting the bullet impacts in silhouettes that integrates image processing and deep learning algorithms for performing pre-processing, impact detection, edge detection, and performance evaluation. A total of 600 silhouettes were created for this project, with bullet impacts of 2401 images suggested and built mask R-CNN system by utilizing the reset 50 and 101 modes revealing that the reset 50 model outperforms the 101 models, with accuracy of 97.6%, 99.5%, and 97.9% respectively. Furthermore, the data shows that both the edge detection and the number of bullet strikes detected are 100% accurate.