CAJ | 학술논문

为了快速、准确、完整地识别裂缝，基于1 mm／像素的路面三维图像提出了具有并行结构的裂缝自动识别算法．首先，进行降维处理，分别以像素（0，0）和（4，4）为起点将源图像划分为8×8像素的子块，获得2幅部分重叠的降维图像；然后，基于降维图像进行裂缝种子识别和裂缝连接，形成10个并列的子流程，从而产生10幅初步裂缝图像；最后，对10幅图像进行裂缝融合与滑动窗口去噪处理，获得裂缝图像．测试结果表明：提出的算法具有较高的准确率（平均92.56％）和召回率（平均90.59％），并以90.59％的 F 值优于 Otsu 阈值分割及 Canny 边缘检测算法；该算法的并行结构有利于程序并行化，能有效提高运算速度．
위료쾌속、준학、완정지식별렬봉，기우1 mm／상소적로면삼유도상제출료구유병행결구적렬봉자동식별산법．수선，진행강유처리，분별이상소（0，0）화（4，4）위기점장원도상화분위8×8상소적자괴，획득2폭부분중첩적강유도상；연후，기우강유도상진행렬봉충자식별화렬봉련접，형성10개병렬적자류정，종이산생10폭초보렬봉도상；최후，대10폭도상진행렬봉융합여활동창구거조처리，획득렬봉도상．측시결과표명：제출적산법구유교고적준학솔（평균92.56％）화소회솔（평균90.59％），병이90.59％적 F 치우우 Otsu 역치분할급 Canny 변연검측산법；해산법적병행결구유리우정서병행화，능유효제고운산속도．
In order to detect pavement cracking rapidly,accurately and completely,an automatic cracking recognition algorithm with a parallel structure is proposed based on 1 mm/pixel 3D pave-ment images.First,image dimensional reduction is conducted.A source image is divided into blocks of 8 ×8 pixels from origin pixels (0,0)and (4,4),respectively,and two partly overlapped images with lower dimensions are obtained correspondingly.Then,crack seed recognition and crack connection are conducted on the two lower-dimensional images,forming 10 parallel sub-workflows, from which 10 preliminary crack images are generated.Finally,the 10 preliminary crack images are fused and then processed via sliding-window denoising techniques,yielding final crack image.Test results show that the proposed algorithm achieves relatively high precision (averaging 92.56%)and recall (averaging 90.59%).It outperforms Otsu threshold segmentation and Canny edge detection with an F score of 90.59%.Furthermore,the parallel structure of the proposed algorithm helps par-allel programming,which can effectively improve computing speed.