In crack images, the distribution of crack pixels and non-crack pixels is very uneven. Therefore, the crack image contains many essential small scale details, as shown in Figure 1. The crack structure change with their mutual connection and multidirectional expansion. These powerful neural network models are an ideal way to aggregate multiple level features. Recently, supervised deep learning methods have achieved state-of-the-art performance in many advanced computer vision tasks, such as image classification, object detection, semantic segmentation and data processing. The shortcomings of traditional crack segmentation methods are clear: once the noise of the scene changes, the effect of the crack segmentation model will become worse. In traditional methods, each crack segmentation method establishes a specific model for different noises. Therefore, it is difficult to design a general method to deal with cracks in various situations.Ĭracks have the following characteristics: complex texture similarity, high inhomogeneity and topological complexity. Uncontrollable factors such as complex road conditions, vehicle occlusion, shadow and weather changes will affect the accurate segmentation of cracks. In addition, due to the complexity of the shooting environment, the low contrast between the crack and the surrounding pixels will also increase the difficulty of detection. The task of road crack segmentation is still facing great challenges due to the complexity of real road conditions and the variability of the data acquisition process. The latter one extracts the crack curve from the overall view by encoding and decoding, and then optimizes the objective function based on certain standards. The former one utilizes local features, such as gradient, texture and variance, to achieve crack segmentation. In general, these methods can be divided into two categories: methods based on local features and methods based on global features. To tackle these problems, some researchers have proposed the use of digital image processing technology, artificial intelligence technology and other methods to effectively detect cracks from images. ![]() In addition, the coverage of manual inspection is limited, and assessment personnel need to investigate roads where vehicles move at high speed, which not only has high security risks but also leads to high inspection costs. Traditionally, road surface defects are evaluated and classified by manual investigation, but this generally depends on, to a great extent, the level of assessment personnel. Human visual inspection is a critical step to assess pavement conditions and for maintaining road safety. Therefore, it is of great importance to detect and analyse cracks for road maintenance and safety. Moreover, road collapse can even cause life-threatening conditions to the surrounding people because of negligence. Pavement faults and defects can reduce the service life of roads if timely inspection and maintenance are not conducted. The experimental results demonstrate that this method achieves state-of-the-art performance on the four challenging datasets.Ĭracks, as one of the most common pavement defects, are a direct reflection of road safety and durability. CurSeg is evaluated on four datasets to validate the effectiveness of the approach. Residual detail attention (RDA) is also introduced to better capture the line structure and the ability to accurately locate the crack position in a complex context to make the network more discriminative and robust. The elaborately designed model can effectively suppress the propagation of noise and further refine the crack features by aggregating multiscale and multilevel features from low-level to high-level. In this approach, features at different scales are fused together to attain the context information from the cracks. In this paper, a deep convolutional neural network called CurSeg is proposed, which achieves pixelwise segmentation of cracks in an end-to-end manner. IET Generation, Transmission & DistributionĪutomatic crack detection is challenging due to the poor continuity of cracks, the different widths of cracks, and the low contrast between cracks and the surrounding pavement.IET Electrical Systems in Transportation.IET Cyber-Physical Systems: Theory & Applications.IET Collaborative Intelligent Manufacturing. ![]()
0 Comments
Leave a Reply. |