关键词: 桩护壁, 裂缝检测, 机器视觉, Canny算子, 相对阈值

Abstract: A crack detection algorithm for protective wall based on machine vision for a range of issues is presented in this paper, such as complex texture and low contrast in surface image of protective wall, unfixed image shooting angle and distance causing relative threshold to be unusable. Firstly, the algorithm enhances the crack contrast through a pixel complement algorithm. Then the distributions of phase angles and gray levels are used to improve Canny operator for restraining the interfering edge points in the image background. Secondly, some multiple relative thresholds to eliminate complicated noise are set up according to the characteristics of crack. Finally, by forecasting the growing trend of the crack, the crack region has been complemented. The experimental results show that the algorithm can preferably detect and identify the crack of protective wall, and also can be applied to the crack detection in the surface of the traditional architectures.

Key words: protective wall for piles, crack detection, machine vision, Canny operator, relative threshold