At present, the application of deep learning algorithms in two-dimensional color image detection is being continuously innovated and broken. With the popularity of depth cameras, color image detection methods with depth information need to be upgraded. To solve this problem, a multi-target detection algorithm based on 3D DSF R-CNN (Double Stream Faster R-CNN, Convolution Neural Network based on Candidate Region) is proposed in this paper. The RGB information and the depth information of the image are given to two input elements of the convolution network with the same structure and weight sharing, and an optimal fusion weight algorithm is used to determine the weight of the fusion target in accordance with the recognition accuracy of the recognition targets under the single modal information, so as to ensure the most efficient fusion result. After several convolution operations, the independent features are extracted and the two networks are fused according to the optimal weights in the convolution layer. With the conducting of convolution and extract the fused features, and finally get the output through the full link layer. Compared with the previous two-dimensional convolution network algorithm, this algorithm improves the detection rate and success rate while ensuring the detection time. The experimental result shows that this method has strong robustness for complex illumination and partial occlusion, and has excellent detection results under non-restrictive conditions.

Abstract:As for the Encryption protection of 3D digital model, 3D digital watermark technology is inevitably the preferred plan. However, due to the curved surface property and three-dimension property in 3D model itself, there are no better plans that can effectively defend all kinds of watermark attacks among the numerous 3D digital watermark algorithms so far. Thus in this paper a modified algorithm will be proposed, which uses a set of complete system of Legendre orthogonal function to conduct the transformative processing to the geometry information of the digital model and embed the watermark into it. The system is a normal complete orthogonal function constituted by Legendre polynomial of times which not only shares the same property with the smooth orthogonal function system but also has the same property of discontinuous orthogonal function system. Therefore, from the result of the test we can see that the algorithm has strong invisibility, stability and robustness, which greatly enhances the security and the reducibility of 3D encrypted mode.

Single biometric identification becomes the bottleneck problem of biometric identification technology inevitably because of its limitation and objectivity. In order to solve the problem, double biometric identification technology will become the main direction of future development. This article is based on this idea and adopts handprints recognition which combines fingerprints recognition with palmprints recognition. It aimed at the complexity and diversity of handprints image and brought forward double biometric identification technology which is based on the handprints. The shortage of fingerprints recognition or palmprints recognition has been overcomed greatly by the methods of Gabor and Wave-let Zerocrossing. Multi-biometric Identification not only improves the efficiency of identification, but also guarantees the accuracy and security of the identification's outcome.

With the development of computer and network technique, no-paper examination has received widely application, such as driving license exam and professional ranks and titles computer exam nearby us. In order to ensure the equity and justness, the issue of organizing papers causes more and more attention. Intelligent Organizing Papers is a combination problem under various constraints. There are still many unideal aspects of the effect, though the technique of organizing papers has high speed development. This thesis focuses on the application of genetic algorithm in organizing papers, and proposes an improved genetic algorithm which increases the efficiency and quality of organizing papers. Keywords-chaotic encryption; digital watermark;DWTWith the widely development of computer, network and communication technique, especially the popularization of Internet, the problem of protection of information security causes more and more attention. Informational exchange based on computers and network brings convenient conditions to the usage and spread of digital media works. Due to the feature of easy illegal copy and misrepresent, digital works' protection is the practical problem urgently needing to be solved.

Recently, a problem that the infrared decoy interferes infrared detection system, cannot be solved. With the gradual application and popularity of the particle swarm optimization, it is preferable to apply it to dynamic multiobjective optimization to solve the problem of the recognition and the estimation of dynamic multi-object in infrared imaging. In this study, the dynamic multi-objective estimation and recognition algorithm of the infrared imaging, which is based on the multi-particle swarms collaboration, ultimately estimates the motion trajectory and Pareto optimal solution of the infrared imaging through the continuous improvement and upgradation of the particle swarms optimized algorithm, the continuous study and inheritance as well as the combination with the aerodynamic characteristic of the infrared decoy. The experiment proves that the improvement of particle swarm algorithm efficiently reduces the estimation error, which produces favorable optimized effects. The experiment has great engineering significance.