Detecting cracks on the concrete surface is crucial for the tunnel health monitoring and maintenance of Chinese transport facilities, since it is closely related with the structural health and reliability. The automated and efficient tunnel crack detection recently has attracted more research studies, particularly cheap availability of digital cameras makes this issue easier. However, it is still a challenging task due to concrete blebs, stains, and illumination over the concrete surface. This paper presents an efficient crack detection method in the tunnel concrete structure based on digital image processing and deep learning. Three contributions of the paper are summarized as follows. First, we collect and annotate a tunnel crack dataset including three kinds of common cracks that might benefit the research in the field. Second, we propose a new coarse-to-fine crack detection method using improved preprocessing, coarse crack region localization and classification, and fine crack edge detection. Third, we introduce a faster region convolutional neural network to develop a coarse crack region localization and classification, then deploy edge extraction to implement the fine crack edge detection, gaining a high-efficiency and high-accuracy performance.deep learning, edge detection, region localization, tunnel crack detection | INTRODUCTIONIn recent years, with the rapid development of road traffic in China, safety inspection of concrete structures is crucial to the cost-effective maintenance of Chinese transport facilities, since it is closely related with the structural health and reliability. As an important part of road traffic, tunnels are also the top priority for security inspection and maintenance. Due to the geology, topography, climate, and construction conditions, lining crack is one of the most common diseases on the surface of tunnel's concrete structure along which it has split without breaking apart.The traditional manual inspection approaches are subjective and time-consuming. Nowadays, more and more researchers make studies on the industrial sensors, such as charge coupled device cameras, ground penetrating radar, and lasers, employ them to catch the crack image and measure the concrete crack automatically. These sensors, especially the cheap cameras, can rapidly collect information over wide concrete areas on the tunnel and get the crack detection results. However, the automatic and efficient tunnel crack detection is still a challenging task due to the concrete blebs, stains, and illumination. This paper focus on judging the lining cracks by acquiring and processing the concrete's image, aiming to reducing the cost of maintenance and time-consuming in the safety inspection of concrete structures.

Detecting the underground disease is very crucial for the roadbed health monitoring and maintenance of transport facilities, since it is very closely related to the structural health and reliability with the rapid development of road traffic. Ground penetrating radar (GPR) is widely used to detect road and underground diseases. However, it is still a challenging task due to data access anywhere, transmission security and data processing on cloud. Cloud computing can provide scalable and powerful technologies for large-scale storage, processing and dissemination of GPR data. Combined with cloud computing and radar detection technology, it is possible to locate the underground disease quickly and accurately. This paper deploys the framework of a ground disease detection system based on cloud computing and proposes an attention region convolution neural network for object detection in the GPR images. Experimental results of the precision and recall metrics show that the proposed approach is more efficient than traditional objection detection method in ground disease detection of cloud based system.

Chip layering defects affect the performance of chips and lead to the failure of chips, so chip layering defects detection is an important step in the quality acceptance of chip production. Chip layering defects, which are characterized by insignificant color change in defect area, small defect area and difficult localization, bring challenges to traditional detection. In recent years, deep learning has shown its powerful ability to solve complex problems in computer vision. In this paper, semantic segmentation method is used to study the problem of chip hierarchical defect detection. Dual focus mechanism first applies whiteboard network structure to identify the true hierarchical area. Afterwards the defective layer area and the original map, the layered defect is recognized in the whiteboard attention. Since the contrast of the layered defect is not obvious, the precise layered defect tag extraction is another important factor affecting network performance. Based on the fuzzy-c-mean clustering algorithm and expert acceptance principle, obtaining the precise layered defect label, the practicality of this method is further enhanced. The effectiveness of the method for detecting the chip layering defects is verified by testing the chip image provided by Huawei.