Deep learning provides an efficient automated method for pavement condition surveys, but the datasets used for this model are usually images taken in good lighting conditions. If images are taken at night, this model cannot work effectively. This paper proposes a method for normalizing pavement images at night, which includes three main steps. First, the image feature point detection and matching method is used to process images taken during the day and night. Then, paired images of pavement during the day and night are obtained. Second, with the help of the image-to-image translation model, those paired images are used for training, and the best model for converting night images into day images is selected. Third, a convolutional neural network (CNN) based on VGGNet is constructed, and pavement images taken during the day are used for training.After that, six types of images are used and tested separately, namely, those taken during the day and the night, converted by the proposed method and converted by traditional methods. As evaluated by various evaluation indices and visualization methods, the detection performance of the CNN model can be significantly improved by using the proposed method of converted night-to-day images.

The three-dimensional (3D) reconstruction technology based on computer vision has greatly facilitated damage inspection and assessment and construction monitoring of civil engineering. However, there are several problems in its implementation. A new method of 3D reconstruction is hereby proposed in this paper regarding the geometrical characteristics of pictures of buildings. An algorithm can normalize the 3D-reconstructed point cloud model so that its length, width, and height are parallel to the coordinate axis of the world coordinate system, and the absolute scale of the point cloud model can be obtained. Compared with the traditional one, this methodology can maintain better accuracy. This paper establishes the theoretical framework of the methodology, and steps for implementation are given by using digital image processing technology. After analyzing the results of a field experiment, it has been proven that this methodology can make the best use of the geometric features and improve the efficiency of the traditional reconstruction algorithm and therefore result in high accuracy.

No abstract

<p>The computer vision algorithm based on deep learning has achieved excellent performance in structural surface damage detection, but the accurate detection algorithm has high requirements for the quantity and quality of data sets. This paper presents a method based on class activation map (CAM), which can detect the crack position and distribution only by image-level data labeling. Firstly, a classification model Vgg16-Crack is trained based on the transfer learning method, and the accuracy and generalization ability of the model are tested by the confusion matrix. Then, based on the CAM algorithm, this paper improves and optimizes the current Grad-CAM++ algorithm, and takes the CAM generated by Vgg16-Crack as the result of crack detection. Finally, the method proposed in this paper is tested in the field. The test result shows that the method proposed in this paper can realize the accurate detection of structural surface cracks.</p>

No abstract