基于计算机视觉的混凝土裂缝识别

Computer Vision-Based Crack Detection and Measurement on Concrete Structure

摘要:

为降低对结构表面进行裂缝识别的经济和时间成本，而采用计算机视觉技术、使用消费级照相机对裂缝图片进行处理，识别裂缝区域和测量裂缝宽度，包括图像模糊、图像增强、形态学运算、图像畸变校准、连通域标记、孤立点消除、裂缝碎片拼接等.对提取出的裂缝区域，统计裂缝发展方向，计算其对应的裂缝长度及宽度.通过钢筋混凝土梁静力加载试验，对梁表面裂缝进行拍摄，在实验室条件下得知裂缝宽度误差在0.1 mm左右.

Abstract:

Crack detection and measurement are accomplished by manual detection with crack scales, which is time-consuming and money-consuming. The research was proposed to detect and measure the cracks on the surface of concrete structure, based on computer vision technology. The measurement device was consumer-grade camera, which was relatively inexpensive. Crack detection involves image blurring, image enhancement, morphological operation, image distortion calibration, connected-domain labelling, isolated point elimination, and crack segmentation matching. Furthermore, the research calculated the length and width, as well as records and development direction of extracted crack pixels. Moreover, in the monotonic loading test of reinforced concrete beams, after taking photos of cracks on the beams and compared with the results by traditional methods, the proposed method attained the decent precision of the computer-vision based crack detection and measurement.