Image Segmentation Method for Complex Vehicle Lights Based on Adaptive Significance Level Set

摘要

The existing study on the image segmentation methods based on the image of vehicle lights is insufficient both at home and abroad, and its segmentation efficiency and accuracy is low as well. On the basis of the analysis of the regional characteristics of vehicle lights and a level set model, an image segmentation method for complex vehicle lights based on adaptive significance level set contour model is proposed in this paper. Adaptive positioning algorithm of the significant initial contour curve based on two-dimensional convex hull is designed to obtain the initial position of evolution curve, thus the adaptive ability of the model is improved. Meanwhile, in order to solve the problem which the image edges are blurred when Gaussian filter is used to remove image noise in Li model, the regularized P-M equation is adopted to achieve effective maintenance of image edge information while the noise is effectively removed. Experimental results show that the image segmentation accuracy for different lights is up to around 95%, the proposed method can significantly reduce the number of iterations and improve segmentation efficiency, which have the advantages of higher accuracy and fast speed, and it can provide a strong support for the accurate vehicle recognition.