基于改进模糊聚类算法的路面裂纹图像分割

图像分割是路面裂纹识别的关键步骤，图像分割的效果直接影响路面裂纹的识别和分类。针对路面图像模糊核均值聚类算法中迭代结果容易出现局部最优的问题。提出一种改进的模糊核均值聚类算法，利用OTSU算法先获得最佳阈值，再通过该阈值得到各聚类的灰度均值，将这些均值作为聚类中心的初始值以实现模糊聚类算法。路面图像裂纹分割试验结果证明，提出的改进算法实现初始聚类中心的优化，避免算法出现局部最优，提高了分割效果，可以应用到路面裂纹图像分割的工程应用中。

Study on Pavement Crack Image Segmentation Based on Improved KFCM Algorithm

The quality of image segmentation directly influences the identification and classification of the pavement crack, and it is the key step for the pavement crack identification. Aiming at the problem of local optimal in the iterative results of the average kernel clustering algorithm for the pavement image blur, this paper put forward an improved fuzzy kernel clustering algorithm, which obtains the best threshold value by the OTSU algorithm, then gets the mean gray value of each clustering based on the threshold value, and makes the mean gray value as the initial value of the clustering center to realize the fuzzy clustering algorithm. The test results of the pavement crack image segmentation show that it realizes the majorization of the initial clustering center, avoids local optimal and improves the segmentation effect. It can be applied to the pavement crack image segmentation.