In order to improve the efficiency of detecting defects on the metal surface of refrigerators and to deal with complex production situations, the Metal-YOLOv3 model was proposed. Using random parameter transformation, the defect data was expanded hundreds of times; the loss function of the original YOLOv3 (You Only Look Once version 3) model was changed, and the Complete Intersection-over-Union (CIoU) loss function based on CIoU was introduced; the threshold of non-maximum suppression algorithm was reduced by using the distribution characteristics of defects; and the anchor value that is more suitable for the data characteristics was calculated based on K-means clustering algorithm, so as to improve the detection accuracy. After a series of experiments, it is found that the Metal-YOLOv3 model is far better than the mainstream Regional Convolutional Neural Network (R-CNN) model in term of detection speed with the Frames Per Second (FPS) reached 7.59, which is 14 times faster than that of Faster R-CNN, and has the Average Precision (AP) reached 88.96%, which is 11.33 percentage points higher than Faster R-CNN, showing the good robustness and generalization performance of the proposed model. It can be seen that this method is effective and can be practically applied to the production of metal products.

Focusing on the issue that releasing frequent sequence patterns and the corresponding true supports may reveal the individuals' privacy when the data set contains sensitive information, a Differential Private Frequent Sequence Mining (DP-FSM) algorithm was proposed. Downward closure property was used to generate a candidate set of sequence patterns, smart truncating based technique was used to sample frequent patterns in the candidate set, and geometric mechanism was utilized to perturb the true supports of each sampled pattern. In addition, to improve the usability of the results, a threshold modification method was proposed to reduce truncation error and propagation error in mining process. The theoretical analysis show that the proposed method is ε-differentially private. The experimental results demonstrate that the proposed method has lower False Negative Rate (FNR) and Relative Support Error (RSE) than that of the comparison algorithm named PFS², thus effectively improving the accuracy of mining results.

Concerning the problems of complicated structure and revolution of 2D traditional neighborhood cellular automata, low encrypting efficiency, little key space of 1D cellular automata, low diffusion speed and needing multiple rounds iteration to produce avalanche effect, a new encryption algorithm based on 2D X-type reversible cellular automata and Arnold transformation was proposed. Firstly, the plaintext was evolved by the proposed cellular automata, then it was transformed by Arnold transformation and cyclic shift transformation after every evolution, until the ciphertext was encrypted well enough. The experimental result shows that the key space is increased by 16.8% and has perfect robustness in resisting brute force attack. In addition the diffusion and confusion is so excellent that it can produce higher avalanche effect and resist chosen plaintext attack.

In order to meet the need of low-cost digital video processing chip,the paper presented a new edge preserving enlargement algorithm using two line buffers．First, a representation point was found to present interpolation point. Next, relevant direction of the representation point was found. Then, four points of this direction and new position relationship were calculated. Last, interpolation algorithm was used. The experimental results demonstrate that the algorithm can not only preserve the edge effectively，but also remove the edge blur． It can be used in the low-cost chip of digital video processing．