2D Gun's type classification using edge detection algorithm and SUSAN low level image processing
V. Khongpit, U. Phangphol, Kulthon Kasemsan -, Chetneti Srisa-An · International Conference for Internet Technology and Secured Transactions · 2012
The objective of this research was to develop the system of gun type classification using Image Recognition Theory. The amounts of 505 images were used as the samples of the study: 80 control, or master, images and 425 experimental images. There were eight gun types using in this experimental which are revolver gun, semi-automatic gun, shotgun, submachine gun, light machine gun, heavy machine gun, recoilless gun and rifle gun. The image samples used in the study were not limited by the image size nor by the direction of the muzzle. The gun types and models classification using image recognition theory (GTMC) comprises of two processes: master image storage process and image recognition process. The preprocessing image is applied to all image samples to adjust the photo scale and the gray scale. Later, Canny was utilized to find the edge of the image. Together with that, SUSAN (Smallest Univalue Segment Assimilating Nucleus) was also applied to trace the edge line and to detect the edge angle of the image. The received data were then examined using block matching between the master image blocks and the experimental image blocks with block matching algorithm. The values received from matching the blocks were the match point and the average similarity of the experimental images against the master images stored in the database. The highest value of the average similarity would be shown as the gun type. It was found from the experiment that GTMC was able to classify the images of the semi-automatic gun with the highest accuracy or 99.06%, and the average accurate gun image classification was 81.25%.