论文部分内容阅读
为了解决在公路路面裂缝的识别和检测过程中,边界细节不明显,噪声干扰不易去除,且存在断裂现象等难点问题,提出了一种新的路面裂缝提取方法。首先利用灰度的垂直和水平投影曲线判断裂缝的类型为横向裂缝、纵向裂缝或网状裂缝;继而根据裂缝的走向进行图像的非对称缩小,减少了背景的无用信息,保留了目标特征边缘,同时提高了检测速度;再采用改进分形理论中的差分计盒维数法对灰度图像分割,即通过计算尺度为1时像素点2×2区域的分维数作为裂缝分割的基础;最后提出通过最大熵阈值法确定阈值,并从端点处循环进行8邻域搜索,找到下个端点,进行断点连接,并去除细小孤立分支,准确地提取裂缝。最后与传统方法进行了对比试验。结果表明:对于具有单条或几条裂缝的图像,新方法能令人满意地提取路面裂缝,抗噪性能好,其结果与人工识别的结果较接近。
In order to solve the difficulties in the identification and detection of the road pavement cracks, the boundary details are not obvious, the noise disturbance is not easy to remove, and the fracture phenomenon exists, a new method of pavement crack extraction is proposed. Firstly, the vertical and horizontal projection curves of grayscale are used to determine the types of cracks as transverse cracks, longitudinal cracks or reticular cracks. Secondly, the image is asymmetrically reduced according to the fracture direction, which reduces the useless information of the background and preserves the edges of target features. At the same time, the detection speed is improved at the same time. Then, the gray-level image is divided by the differential box dimensionality method in the improved fractal theory, that is, the fractal dimension of 2 × 2 pixel area is calculated as the basis of fracture segmentation. The maximum entropy threshold method is used to determine the threshold value, and a loop of 8 neighborhood search is carried out from the end point. The next endpoint is found, the breakpoint connection is made, the fine isolated branch is removed, and the crack is accurately extracted. Finally, compared with the traditional method. The results show that the new method can extract pavement cracks satisfactorily with single or several cracks, and has good anti-noise performance. The result is close to the result of artificial recognition.