Vehicle visual navigation algorithm based on detect of horizon

Combined with the demand of vehicle's visual navigation system and the features of aerial images, a horizon detection algorithm which is used by visual navigation and based on the similarity between regions of images and tabu search was proposed. A single frame extracted from aerial video sequences was delt with downsampling and noise reduction. Tabu search algorithm parameters, such as initial solution, were initialized. Tabu search algorithm combined with the fit function fit the position of horizon. Tabu search's amnesty and convergence criteria were used to judge the value of fit function until it meet convergence criteria. The values of the bank angle and the pitch angle which decide the next parameter-adjusting interval, and send the results to control system were driven. Compared with the traditional visual navigation method and the methods based on Hough transform, the experimental results show that the proposed algorithm is no-reference, good real-time and robust.