Abstract:As the common defects, cracks are critical to the performance of civil
engineering structures. With the continuous optimization of inspection system based
on machine vision, the effective crack recognition and statistics are developed and
become the key of the whole system, in which selection of characteristics is
involved. After preliminary characteristics selection, the reasonable distribution
and utilization of the characteristics should be further studied. After selecting
the crack characteristics, the validity of crack characteristics application is
verified by filtering a large number of impurities from the processed binary
images. And then, an automated identification model is established according to
the RBF-SVM algorithm. 6 characteristics (all) and 3 characteristics (part) are
used as the input parameters, respectively. It shows that this model has good
adaptability and high-efficiency in crack identification, which further validates
the necessity of the distribution and utilization of crack characteristics.