Efficient detection of pavement cracks can effectively prevent traffic accidents and reduce pavement maintenance costs. In order to overcome the complicated and uneconomical disadvantages of traditional crack detection methods, this paper introduces a pavement crack detection network based on deep learning, which can automatically detect pavement cracks and achieves excellent detection accuracy. And the network can easily use the sensors to collect data to facilitate industrial applications. In additional, considering that most cracks have slim feature, we apply the latest Transformer module in the network to improve the effect of cracks detection. Transformer has a strong ability to capture the long-range dependence of the cracks, which enables the network to learn the context information of the crack region. Furthermore, the network also utilizes some techniques to improve the ability of algorithm to detect various cracks. Our network is trained on pavement data sets containing India, the Czech Republic and Japan. It achieved F1 scores of 0.6739 and 0.6650 on two online test sets with fewer network parameters.

DNA computing based image encryption is a new, promising field. In this paper, we propose a novel image encryption scheme based on DNA encoding and spatiotemporal chaos. In particular, after the plain image is primarily diffused with the bitwise Exclusive-OR operation, the DNA mapping rule is introduced to encode the diffused image. In order to enhance the encryption, the spatiotemporal chaotic system is used to confuse the rows and columns of the DNA encoded image. The experiments demonstrate that the proposed encryption algorithm is of high key sensitivity and large key space, and it can resist brute-force attack, entropy attack, differential attack, chosen-plaintext attack, known-plaintext attack and statistical attack.

The edge-based active contour model has been one of the most influential models in image segmentation, in which the level set method is usually used to minimize the active contour energy function and then find the desired contour. However, for infrared thermal pedestrian images, the traditional level set-based method that utilizes the gradient information as edge indicator function fails to provide the satisfactory boundary of the target. That is due to the poorly defined boundaries and the intensity inhomogeneity. Therefore, we propose a novel level set-based thermal infrared image segmentation method that is able to deal with the above problems. Specifically, we firstly explore the one-bit transform convolution kernel and define a soft mark, from which the target boundary is enhanced. Then we propose a weight function to adaptively adjust the intensity of the infrared image so as to reduce the intensity inhomogeneity. In the level set formulation, those processes can adaptively adjust the edge indicator function, from which the evolving curve will stop at the target boundary. We conduct the experiments on benchmark infrared pedestrian images and compare our introduced method with the state-of-the-art approaches to demonstrate the excellent performance of the proposed method.

When the traditional semantic segmentation model is adopted, the different feature importance of feature maps is ignored in the feature extraction stage, which results in the detail loss, and affects the segmentation effect. In this paper, we propose a BiSeNet-oriented context attention model for image semantic segmentation. In the BiSeNet, the spatial path is utilized to extract more low-level features to solve the problem of information loss in deep network layers. Context attention mechanism is used to mine high-level implied semantic features of images. Meanwhile, the focus loss is used as the loss function to improve the final segmentation effect by reducing the internal weighting. Finally, we conduct experiments on open data sets, and the results show that pixel accuracy, average pixel accuracy, and average Intersection-over-Union are greatly improved compared with other state-of-theart semantic segmentation models. It effectively improves the accuracy of feature extraction, reduces the loss of feature details, and improves the final segmentation effect.