Exploiting bilinear interpolation and predictive particle swarm optimisation for tilt correction of vehicle license plate images

Images of tilted vehicle license plates reduce the accuracy of automatic license plate detection and recognition systems. In this paper, we present the analysis for how to calibrate tilt angles of vehicle license plates. We applied an edge detection scheme for tilt correction, which obtains an accurate estimation of the incline angles, where the tilt to either horizontal or vertical lines can be eliminated by rotation using bilinear interpolation and offset correction. Segmentation of characters on license plates is achieved using vertical projection combined with prior constraints. For the license plate recognition system, we also applied an improved chaotic particle swarm optimisation (CPSO) scheme for character recognition. Experimental results show that bilinear interpolation-based approach displayed show that bilinear interpolation-based approach displayed an average of 20% less location time and 55% erroneous rate than those of Sobel edge detector, while the improved CPSO scheme is capable of achieving an average of 15% operation time and 40% iteration time than those of original CPSO for the segmented license plates. Simulations demonstrate the robustness and efficiency of our algorithm.