Defect classification is an important part of steel surface defect detection. When the Convolutional Neural Network (CNN) has achieved good results, the increasing number of network parameters consumes a lot of computing cost, which brings great challenges to the deployment of defect classification tasks on personal computers or low computing power devices. Focusing on the above problem, a novel lightweight network model named Mix-Fusion was proposed. Firstly, two operations of group convolution and channel-shuffle were used to reduce the computational cost while maintaining the accuracy. Secondly, a narrow feature mapping was used to fuse and encode the information between the groups, and the generated features were combined with the original network, so as to effectively solve the problem that "sparse connection" convolution hindered the information exchange between the groups. Finally, a new type of Mixed depthwise Convolution (MixConv) was used to replace the traditional DepthWise Convolution (DWConv) to further improve the performance of the model. Experimental results on NEU-CLS dataset show that, the number of floating-point operations and classification accuracy of Mix-Fusion network in defect classification task is 43.4 Million FLoating-point Operations Per second (MFLOPs) and 98.61% respectively. Compared to the networks of ShuffleNetV2 and MobileNetV2, the proposed Mix-Fusion network reduces the model parameters and compresses the model size effectively, as well as obtains the better classification accuracy.

Traditional edge detection methods can not accurately detect the Virtual Boundary (VB) between different regions in materials microscopic images. In order to solve this problem, a virtual boundary detection model based on Convolutional Neural Network (CNN) called Virtual Boundary Net (VBN) was proposed. The VGGNet (Visual Geometry Group Net) deep learning model was simplified, and dropout and Adam algorithms were applied in the training process. An image patch centered on each pixel in the image was extracted as the input, and the class of the image patch was output to decide whether the center pixel belongs to the virtual boundary or not. In the experiment of virtual boundary detection for two kinds of material images, the average detection precision of this method reached 92.5%, and the average recall rate reached 89.5%. The experimental results prove that the VBN can detect the virtual boundary in the image accurately and effectively, which is an alternative method to low-efficient manual analysis.

To deal with the lung segmentation problem under complex conditions, Random Walk algorithm was applied to automatic lung segmentation. Firstly, according to the anatomical and imaging characteristics of the chest Computed Tomography (CT) images, foreground and background seeds were selected respectively. Then, CT image was segmented roughly by using the Random Walk algorithm and the approximate mask of lung area was extracted. Next, through implementing mathematical morphology operations to the mask, foreground and background seeds were further adjusted to adapt to the actually complicated situations. Finally, the fine segmentation of lung parenchyma for chest CT image was implemented by using the Random Walk algorithm again. The experimental results demonstrate that, compared with the gold standard, the Mean Absolute Distance (MAD) is 0.44±0.13 mm, the Dice Coefficient (DC) is 99.21%±0.38%. Compared with the other lung segmentation methods, the proposed method are significantly improved in accuracy of segmentation. The experimental results show that the proposed method can solve the difficult cases of the lung segmentation, and ensure the integrity, accuracy, real-time and robustness of the segmentation. Meanwhile, the results and time of the proposed method can meet the clinical needs.

In allusion to the problem that how to detect whether the digital image data is complete, and whether the image is tampered, an image authentication algorithm based on two-dimensional histogram shifting was proposed to improve the quality of the authentication image. Firstly, the two-dimensional histogram of the cover image was structured through two prediction difference value calculating methods. The embeddable channels were chosen by the preset parameters and the embeddable channel peak positions were determined and the embeddable channels were shifted. Then, the anthentication information was embedded into image blocks with histogram shifting method. Hierarchical tampering detection was adopted during the process of tamper detection to effectively improve the accuracy. The experimental results showed that the algorithm could resist noise attack, and the average Peak Signal-to-Noise Ratio (PSNR) of the authentication image was 52.37dB and 50.33dB respectively when the parameter was set as 2 and 4, which improves the quality of images. The resutls prove that the algorithm has high security, and it is able to implement reversible watermarking as well as precisely locationing the tampering region.

Aiming at the problem of pulmonary small nodules was difficult to identify, a method using fuzzy C-means clustering algorithm to analyse the lung Region Of Interest (ROI) was presented. An improved Fuzzy C-Means clustering algorithm based on Plurality of Weight (PWFCM) was presented to enhance the accurate rate and speed of small nodules recognition. To improve the convergence, each sample and its features were weighted and a new membership constraint was introduced. The low sensitivity from the uneven ROI data was decreased by using a double clustering strategy. The experimental results tested on the real CT image data show that PWFCM algorithm can detect lung nodules with a higher sensitivity and lower false positive rate.

To overcome the shortcomings of Cross-Media Relevance Model (CMRM) whose efficiency and effectiveness are low, an improved CMRM was proposed. Based on the improved smoothing method for textual words, the improved CMRM simplified the feature representation and similarity computation which made the measure of relationship between image and image more accurate. The experimental results on the Corel5k dataset show that the proposed approach can significantly improve annotation efficiency. The performance of the improved CMRM is almost three times as good (in terms of mean F1-measure) as original CMRM, also, better than some previously published high quality algorithms such as famous Multiple Bernoulli Relevance Model (MBRM) and Supervised Multiclass Labeling (SML).

Considering that contact duration has significant influence on whether packet can be transmitted successfully or not, the authors proposed a Probabilistic Routing Protocol using History of Encounters and Transitivity based on Contact Duration (PRoPHET-CD), which combined contact duration with encounter frequency to estimate delivery probability. This protocol could improve the delivery probability significantly and reduce the interruption of packet transmission. The simulation results show that the protocol of PRoPHET-CD can significantly enhance the message delivery probability and reduce the overhead ratio.

Adaptive Neuro-Fuzzy Inference System(ANFIS) is a neural network system with one input and several outputs, which is got through achieving one step sugeno fuzzy system by the format of network. It can simulate the connection of the in-out very well, and its constringency is very soon, error is very little and the necessary swatch is little, which are very favourable for blind separation of watermark signals.Based on the upstanding adaptive control ability of ANFIS, a strongly robust audio watermarking algorithm in DCT domain was given in this paper. Experiments show that the performance in watermark detection is high, the time expense is litlle, the resisting to attacks is strong and the imperceptibility is good.