Pedestrian detection is a crucial aspect of target detection in computer vision. The main objective of this field is to process input video or image data using deep learning network models and label pedestrian target locations using bounding boxes on the original data. Currently, deep learning-based pedestrian detection methods have become dominant in the field due to the rapid development of convolutional neural networks and computer technology. However, existing target detection algorithms face challenges when performing the task on embedded devices due to their complex network structure, large computational and parametric quantities, and limited energy efficiency ratio and power consumption. Furthermore, the current pedestrian detection model still experiences a high number of missed detections due to the uncertainty of the occlusion pattern and the prevalence of occlusion within the dense crowd class. To tackle the aforementioned issues, this paper presents a lightweight network architecture achieved through innovative enhancements to the YOLOv7 algorithm. The primary research focuses on optimizing the network by introducing the RepGhost module. This involves utilizing the original ELAN module and ELAN-H module from YOLOv7 to construct a lighter feature extraction network. Secondly, following the concept of RepPAN, we introduce a new efficient neck network, named RepFPN, with the goal of enhancing both accuracy and efficiency of the model. Finally, we replace the RepBlock module in RepFPN with the RepGhost module to achieve deep fusion of two lightweight and efficient structures. The experimental results on the CityPersons dataset show that the YOLOv7_RepGFPN_RepGhost network model proposed in this paper, compared with the original model, reduces the number of parameters of the network model from the original 36.90M to 20.52M while the detection accuracy is improved by 0.6% to 68.6%, and the number of processed frames per second is also somewhat improved to 32.9 FPS.

CNN models are prevalent deep learning models utilized in the fields of machine learning. However, their implementation in hardware has encountered challenges, including high computational complexity, large storage requirements, memory bandwidth limitations, and difficulties with parallel computing. This study introduces a dedicated hardware accelerator designed to enhance the performance of convolutional neural networks, leveraging efficient computational units and memory hierarchy to attain accelerated processing. Initially proposed is an efficient and straightforward Img2Col method, through which convolutions can be unfolded into matrix computations from small size 3×3 to more extensive 16×16. A customizable systolic array is subsequently designed to support the acceleration convolutions. Our well-designed accelerator has been implemented utilizing the hardware description language SpinalHDL and tested on the ZYNQ. The experimental results demonstrate that our accelerator showcases remarkable performances in both CNN and GEMM calculations, delivering up to 37.6 GOPS/W.

To study the influence of rainfall and earthquake on slope stability, Based on the slope of the Limin Tunnel entrance section of a high-speed railway from Harbin to Mudanjiang, Slope stability SRM (strength reduction method) in finite element analysis software Midas GTS NX is used, Eigenvalue analysis and nonlinear time history +SRM analysis, The displacement, strain, and safety factor of the slope under natural, rainfall and earthquake conditions are analyzed. The safety factors of the slope under natural, rainfall, and earthquake conditions are 1.200, 1.0250 and 0.9887, respectively. According to GB50330-2013 "Construction slope Engineering Technical Code"  in addition to the natural conditions, the slope stability is poor under the conditions of rainfall and earthquake, so strengthening measures should be taken, In this paper, the most common prestressed anchor in slope reinforcement is adopted. According to the position of the potential sliding surface, the prestressed anchor is adopted to strengthen the silty clay layer and strongly weathered granite layer, And its stability after reinforcement is analyzed. After the reinforcement of prestressed bolts, the No. 2 and No. 3 bolts play an important role,Its largest bolt prestress reached 308,000 KN/, 227,000 KN/.The safety factor under rainfall and earthquake conditions reached 1.465 and 1.387 respectively, and the slope could satisfy the stability requirements under both rainfall and earthquake conditions.

The development of T cells undergoes a series of finely regulated stages and selections in the thymus. This paper aims to delve into the key steps and mechanisms of this process. Through a detailed analysis of the DN, DP, and SP stages of T cell development, as well as discussions on the mechanisms of positive and negative selection, we reveal the complexity of T cell development. Additionally, this paper highlights the structural features, generation process, and gene rearrangement mechanisms of the T cell receptor (TCR), emphasizing the importance of TCR in T cell diversity and immune response.

The handling of materials with the right quantity, the right cost, and the right time is vital for the manufacturing company. This study is about the optimum level of investments that are made in stores, where the quantity and cost of materials should also be optimal. Nowadays, the majority of firms either spend too much money on overstocking or lose customers as a result of understocking. Thus, the issue has been addressed, and there is a need to examine the stocks in the manufacturing company. This study is being carried out based on the past 20 years of demand data at the store of Jeevan Solvent Extracts Private Limited, basically a rice bran oil manufacturing company, in an effort to stop overstocking and understocking of materials, which results in various consequences and expenses. In order to avoid this, stockout cost and optimum level of safety stock can be examined through various analyses, such as finding out the annual daily demand, reorder point, probability of the number of times the quantity was demanded, additional carrying cost, additional stockout cost, and total relevant cost. By examining those analyses, we can identify the optimum level of safety stock, and a new reorder point can be obtained. This will assist the company in determining at which point the minimum level of investments should be made, and this method will also be helpful for all manufacturing companies in figuring out their ideal safety stock level, their optimum reorder point and when to bare the minimum level of investments that are made in stores.

This study focuses on Arogya Garments, a textile leader, to explore how their inventory management practices affect their financial health. Recognizing the importance of efficient inventory control for cost management, customer satisfaction, and overall operations, the study will analyse the connection between inventory practices and economic indicators like profitability and resource utilization. The goal is to provide valuable insights that Arogya Garments can use to optimize its inventory strategies. Ultimately, this study not only helps Arogya Garments improve efficiency and strengthen its market position but also contributes to a wider understanding of how inventory management within the textile industry.

With the rapid development of location-based smart services and vehicular network technologies, the collection and application of traffic trajectory data have demonstrated significant potential in fields such as intelligent driving, traffic management, and data analysis. The extensive utilization of user trajectory information within vehicular networks has highlighted data security and privacy protection as critical issues needing urgent resolution. Particularly, against the backdrop of integrated development of vehicle positioning technologies and self-organizing networks, the research focus has shifted to how traffic trajectory information privacy can be effectively safeguarded without compromising the accuracy of location services. This paper delves into how personalized privacy protection for traffic trajectories can be achieved within location-based service platforms, ensuring data usability. It provides fundamental theories, technical approaches, and evaluation methods to quantify the privacy needs of traffic trajectory data, thereby safeguarding personalized trajectory privacy. Using the PORTO-TAXI dataset as a case study, this research explores the design and application methods for trajectory privacy protection, offering innovative solutions to the field of traffic data privacy protection.