Vehicle License Plates Tilt Detection Based on the p-th Power Variance of Projection Coordinates and Particle Swarm Optimization

The commonly used tilt detection algorithm for license plate based on rotation projection is time consuming besides its insufficiency of detection accuracy.In view of it,a new approach is presented based on the minimum p-th power variance of projection coordinates and particle swarm optimization.Firstly,to further improve detection accuracy,the tilt angle of the license plate is determined based on the minimum p-th power variance of vertical coordinates of the horizontally projected edge points.Then the optimal tilt angle is found by using the extremum disturbed and simple particle swarm optimization algorithm or calculated directly so as to reduce the detection time.Finally,the effects of different p values on the license plate tilt detection error and the running time are experimentally studied and the performance is compared with those of Hough transform method,rotation projection method and principal component analysis method.The experimental results show that,when p→∞,the proposed approach has the highest accuracy,while p=2,the computation is the fastest.The proposed approach has been successfully applied to our "Intelligent Parking" system.