Chinese medicine has a pivotal position in China and the world, and Chinese herbal medicine, as an essential part of it, naturally attracts much attention. The monitoring and management of the supply chain of Chinese herbal medicine is an important link to maintain the healthy development of the Chinese herbal medicine industry. At the present stage, the supply chain supervision of Chinese herbal medicines still has problems such as information asymmetry and difficulty in traceability and accountability, which cannot guarantee the quality and safety of Chinese herbal medicines. In order to improve the quality of safety supervision of Chinese herbal medicines by relevant departments and improve the operational efficiency of the Chinese herbal medicine supply chain, we analyzed the factors affecting the traceability of Chinese herbal medicine supply chain in China and constructed a traceability system of Chinese herbal medicine supply chain using the process trustworthiness and decentralization features of blockchain to provide solutions for the traceability of Chinese herbal medicine quality and safety.

In the era of big data, the development of location sensing technology such as mobile communication and sensing devices has formed location big data, which has brought huge benefits to people's life, business operation methods and scientific research. Due to the diverse uses of location big data and the overlapping redundancy of contents, the classical privacy protection methods based on "informed consent" and anonymity can not fully protect user privacy. The privacy protection technology of location big data measures the user's location privacy and protects the user's sensitive information in the sense of information theory. This paper introduces the concept of location big data and the privacy threats of location big data, summarizes the unified metric-based attack model for location big data privacy, and summarizes the existing research achievements in the field of location big data privacy protection. The user's knowledge of text content and online behavior can be used to carry out reasoning attacks on users. Social relational reasoning and attribute reasoning are two basic attacks on user privacy in social networks. The research on the protection mechanisms and methods of inference attacks is also increasing. This paper classifies and summarizes the research and literature related to privacy inference and protection techniques, and finally discusses and looks into the future.

The Internet provides us with rich data and services. Identity authentication and resource access authorization are key technologies in the field of network services and management. To realize the transmission of multi-domain identity information with only one cross-domain request process, we make full use of the characteristics of distributed consistency, combine the teaching experience of cross-domain identity authentication and data security in recent years, we propose a certificateless cross-domain identity authentication scheme based on master-slave chain. The scheme provides reference for the application teaching reform of data security by discussing the case design of cross-domain identity authentication.

Group photography is a common image in life, but when you take a group photo, there will always be someone who can't be there for a variety of reasons, resulting in a group photo always with regret. Especially in recent years, because of the epidemic, many people at the gathering could not be reached. As a result, the research on the generation of group photo images of people becomes more and more important. The purpose of this study is to perfectly embed a single person image into a multi-person photo image to make it harmonious and realistic. Previous people's image compositions focused more on the fusion of characters and backgrounds, with little regard to the interaction between people and other people in the group photo. Now, it is more desirable to have a good fusion with multi-person photo images. In this survey, we summarize the methods of the above research directions, and discuss the difficulties and future directions of future group photo character image generation to promote future research of group photo realization.

Autoencoder network is a kind of network structure in current deep neural network. In recent years, autoencoder networks have been used in the tasks of dimensionality reduction of various data and feature extraction of complex things. Geodesic distance describes the local shortest path between two points on a spatial manifold, which is an essential measure of the manifold.This paper uses the Swiss roll function as the main research function, combines the auto-encoding network with the geodesic distance, uses the auto-encoding network to input the coordinates of the points in the space as input values     into the auto-encoding network, and extracts and trains the features of the point coordinates. A new set of high-dimensional coordinates is obtained, so that the Euclidean space distance obtained from this set of high-dimensional coordinates can be closer to the geodesic distance in the actual space. The research content of this paper mainly includes the following three parts:

1. The development, concept and theory of auto-encoding network are introduced, and the principle of auto-encoding training is briefly described.
2. The Floyd algorithm is used to solve the geodesic distance between two points of the Swiss roll function, and the approximate geodesic distance is obtained.

Use the auto-encoding network to train and test the coordinates of the points in the Swiss roll function to obtain a new set of coordinates, calculate the Euclidean space distance of the new set of coordinates, and use this set of Euclidean distances with the geodesic distance measured by Floyd Make a comparative estimate and calculate the error.

The study aimed to examine characteristic impacts and new livelihood rehabilitation activities among PAP caused by NN2 hydropower project in Laos. The results of study revealed that, the reasons behind construction of NN2 project were not only for renewable and energy supply, but also included to eradicate insecurity and poor living conditions among rural people. The project had agrarian livelihood impacts among rural villages. Those new livelihood activities were rehabilitated by compensating lost properties, providing social services, establishing vocational groups, providing agricultural area and inputs, practicing on-farm and off-farm activities, increasing annual income, reducing poverty, implementing MDGs and enhancing capacities.

Paper An enhanced YOLOv3 pedestrian recognition model is developed to address the problem of YOLOv3's easy to miss small objects and occlusion problem. The upgraded model replaces the original K-means clustering algorithm with the K-means++ clustering method to lessen the error effect caused by incorrect random selection of starting clustering centers. Meanwhile, the residual network module is introduced to the model to lighten it up, and the CBAM attention mechanism and the MHSA multi-head self-attention mechanism are incorporated into the framework. The algorithm's feature extraction capability is enhanced by efficiently allocating computational resources and gathering global information.Experiments testing the FLIR dataset on NVIDIA Xavier NX show that this al-gorithm greatly reduces the number of network model parameters with less loss of accuracy and improves the FPS.The experiment shows that the revised method performed well on the CUHK dataset, with the mAP value acquired during the experiment being 88.20%, 17.45% higher than the original algorithm. It has greater feature extraction capabilities, increases identification of small targets and blocked pedestrians, and has higher detection accuracy.