Harris corner method for video-based vehicle number plate detection

Vehicle number plate (TNKB) detection is part of a smart city's Intelligent Transportation System (ITS). TNKB is the most easily observed and unique vehicle identity so it can be used as a basis for vehicle identification. In this study, the authors detected and cropped the TNKB position. The process begins with creating a Region of Interest, background subtraction, thresholding, segmentation, and vehicle tracking. In general, background subtraction is formulated as P(i, j) = |A(i, j)-B(i, j)| where P(i, j) is the foreground matrix, A(i, j) is the representation matrix in a frame, and B(i, j) the background matrix. In the next stage, edge detection is performed on the image by calculating the x and y derivatives, and then the author performs dot product operations on each pixel. In the next process, each pixel is convoluted with a gaussian filter to get a response from each pixel. Then calculate the non-max suppression to get the points used as a reference to get the position of the TNKB. The author experimented on four videos of vehicles on the highway, obtaining an average accuracy of 84.36%.