In the current process of social development, reimbursement has become a generally accepted phenomenon. With the improvement of economic level and the improvement of people’s living standards, all walks of life have developed rapidly, which also provides new ideas for the financial reimbursement system and financial management. At present, most of the financial reimbursement processing is conducted manually, which can not meet people’s requirements for work efficiency. Moreover, there are many limitations, which are very unfavorable for enterprises. Therefore, it is necessary to take reasonable and effective measures to strengthen the improvement and optimization of the financial reimbursement system, so as to ensure the safe and efficient operation of funds. Image recognition technology is an indispensable and important means of modern information management. It can automatically extract data information and analyze statistics, which brings great convenience to financial reimbursement. This paper mainly studied the problems related to financial reimbursement based on the process of image recognition and denoising, and put forward some suggestions for the design of financial reimbursement image recognition system. It is hoped that it can promote its better application in practical work, so as to achieve the purpose of improving economic efficiency and ensuring the security of funds, and at the same time help further promote the healthy and orderly development of enterprise construction. This paper compared the traditional manual reimbursement method with the financial reimbursement automatic entry system based on image recognition. The results showed that the error of automatic input system was smaller than that of manual mode, and the degree of automation was higher; in addition, the accuracy rate of reimbursement voucher identification and review had also increased by about 6.34%. Therefore, this method has good advantages and practicability, and this method is conducive to reducing the workload of staff and facilitating the follow-up work. To sum up, electronic imaging technology can analyze and process data with the help of image processing means, thus obtaining corresponding results. It is convenient to adjust the accounting process as needed and timely in the process of financial management, so as to make the overall financial reimbursement work more standardized and unified.

The development of society has led to the continuous development and progress of artificial intelligence technology, and has also led to an increasing demand for graphic design. In order to better solve the problems of color deviation, poor design effect, and high design cost in traditional graphic design, this article applied artificial intelligence image identification system to graphic design to overcome the problems of traditional graphic design. The elements extracted from the graphic database were denoised and enhanced by means of mean filtering and histogram equalization; after image preprocessing, Deep Learning (DL) algorithms were used to construct an image identification system, and the modules and visualization interfaces of the system were introduced. Through experiments, it could be found that the average expert rating of the graphic design scheme designed by the DL based image identification system was 8.818 points, and the satisfaction rate of the 20 users selected for the DL based image identification system was above 93.4%. In summary, using DL to construct an image identification system and applying it to graphic design could effectively improve the overall effect of graphic design and increase user satisfaction with the designed graphic scheme.

Adapt to the new competitive environment, the supply chain concept and management model of horizontal integration and cooperation between enterprises have begun to rise, and continuously demonstrate enormous competitive strength and superiority. However, the existing enterprise supply chain management (SCM) system has problems of low security, low efficiency, and high management costs. In view of the above problems, this paper studied the enterprise supply chain management and its information assurance mechanism based on the error back propagation algorithm. By analyzing the problems in enterprise supply chain management and introducing error back propagation algorithm as an optimization method, the efficiency and accuracy of the supply chain have been improved. At the same time, corresponding guarantee mechanisms were proposed to address the importance of information security in the enterprise supply chain. The research results indicated that the information leakage rate of the supply chain information protection mechanism based on the error back propagation algorithm was below 3.21%, and the average leakage rate of 20 experiments was 2.654%. For supplier management in enterprise supply chain management systems, the selected users scored the system based on error back propagation algorithm at least 8.84 points, and the average score of 10 users was 8.995 points. Enterprise supply chain management and information assurance mechanism based on error back propagation algorithm can effectively improve the effect of supply chain management and enhance the security of information.

Karst water plays a vital role in meeting daily population needs. Determining groundwater sources, understanding chemical changes, and accurately evaluating flow paths and evolution stages are essential for the protection and sustainable use of groundwater resources in mining areas.This study collected 10 sets of karst groundwater and surface water samples from the Anle Village mining area. Using multivariate statistical analysis, Piper trilinear diagrams, Gibbs diagrams, and isotopic techniques, we analyzed the hydrogeochemical characteristics of both contaminated and uncontaminated water samples.The results show that uncontaminated groundwater and surface water are slightly alkaline and dominated by Ca²⁺ and Mg²⁺ cations, along with HCO₃⁻ and SO₄²⁻ anions. Hydrochemical facies include HCO₃⁻-SO₄²⁻-Ca²⁺-Mg²⁺ and HCO₃⁻-Ca²⁺-Mg²⁺.Uncontaminated samples contain high levels of impurities, with dominance of Ca²⁺, Mg²⁺, and SO₄²⁻. These waters are mainly recharged by atmospheric precipitation and influenced by evaporation. Their chemical composition is primarily driven by the weathering and dissolution of carbonate, sulfate, and silicate rocks.Nitrate (NO₃⁻) concentrations in surface water suggest influence from agricultural fertilizers, while contaminated groundwater is closely linked to mineral resource development.These findings are significant for understanding the circulation and evolution of karst water in Anle Village and for informing the protection and utilization of local water resources.

As an indicator of climate change, the change of vegetation cover directly reflects the ecosystem dynamics of the region. In this paper, the spatial and temporal characteristics of vegetation cover in the headwaters of the Fen River and the effects of temperature, precipitation, GDP and population on the changes of vegetation cover were statistically analyzed by using the Theil-Sen median slope and the Mann-Kendall test and Pearson’s correlation coefficient from 2000 to 2020. The results showed that: (1) from 2000 to 2020, the vegetation cover of the Fen River headwaters showed an overall upward trend, and the mean value of NDVI was 0.55. The fluctuation increased from 2000 to 2011; the significant increase was observed from 2011 to 2013; and the fluctuation of the value of NDVI from 2013 to 2020 was relatively small   p  0.01 . (2) Climate change affects changes in vegetation cover. On the time scale, the 2000-2020 mean NDVI values are positively correlated with temperature and precipitation, but the correlation is not significant   p p     0.053 0.05, 0.185 0.05 . On the spatial scale, vegetation cover was weakly negatively correlated with air temperature as a whole, while positively correlated with precipitation as a whole. (3) The influence of human activities on vegetation cover was dominant, NDVI and GDP were positively correlated, with only 5.13% negatively correlated in the central and northeastern part of the region, and NDVI and population were strongly positively and negatively correlated, with alternating distribution in the study area. (4) The vegetation cover of the Fen River headwaters area shows an increasing trend, but there are still ecological and environmental problems, and it is necessary to continue to improve the implementation of the relevant ecological protection policies in order to achieve the goal of sustainable development. The results of the study can provide scientific references for the restoration of vegetation cover and protection of fragile ecosystems in the transition zone of semi-arid and semi-humid climate.

In enterprise cost accounting and control research, traditional activity-based costing (ABC) relies on detailed activity tracking data and complex cost allocation models, which makes data acquisition difficult, has low-cost allocation accuracy, ignores dynamic changes, and has the problem of insufficient flexibility. This paper constructs an improved ABC application framework, builds an activity-driven cost accounting model, analyzes the daily activity data of the enterprise, determines the key factors related to cost, and establishes a mapping relationship between activity and cost. This paper introduces a dynamic adjustment mechanism to adjust the weights and parameters in the cost accounting model in real time according to changes in the external environment and internal operations, thereby improving the flexibility and accuracy of cost accounting. It can integrate the ERP (Enterprise Resource Planning) system with the cost accounting model, integrate the company’s financial data, production data and sales data, use information tools to automatically update activity costs, and provide timely feedback to the cost control system; it can closely combine cost accounting and control, monitor and adjust costs in real time during the accounting process, and take timely control measures when abnormalities occur. Experiments show that in terms of cost allocation accuracy, the average SE (Standard Error) of the improved ABC in enterprises with different employee sizes is 2.1, and the average MSE (Mean Squared Error) is about 5.5. It is more stable when processing enterprise data and can better reflect the actual cost allocation. The response time of the improved ABC is 5.7 seconds when the raw material price increases by 25%. It can make adjustments faster, with better flexibility and dynamic adaptability; the experiment proves the effectiveness of this paper in the research of enterprise cost accounting and control.

In response to the shortcomings of traditional enterprise financial management information platforms in data processing and analysis efficiency and decision support capabilities, this study introduces intelligent decision support systems to fundamentally improve these issues. In this study, we automated data collection through API (Application Programming Interface) technology, used ETL (Extract, Transform, Load) tool for data format conversion, and strictly performed data cleaning and standardization to ensure data quality. The article uses association rules and support vector machine machine learning algorithms for in-depth analysis and prediction of financial data, and optimizes decision-making scenarios based on multi-criteria decision analysis, Monte Carlo simulation and linear programming techniques. Evaluation results show that the system significantly improves the speed and accuracy of data processing, with an increase in processing efficiency of more than 70% and a decision-making accuracy rate of up to 95%. The intelligent decision support system effectively improves the informatization level of enterprise financial management and provides more scientific and reliable decision support for the enterprise leadership.

Focused crawlers are targeted to search the internet for web pages on specific topics. Its main task is to collect preprocessed and topic related web pages and ignore irrelevant web pages. Traditional focused crawlers have limited success in achieving multi-text categorization of web pages. Due to the large amount of unstructured data present in web pages, the correct classification of web pages based on a given topic is the main practical challenge for focused crawlers.The main objective of this work is to design an improved focused crawling approach using web page classification. In this paper, a text classification model based on the combination of GloVe word vector model and TF-IDF weighting technique is proposed to improve the accuracy of web page classification. The GloVe-based text classification model is further utilized to guide focused crawlers to classify web pages.The proposed GloVe and TF-IDF text categorization models are validated on 10 different datasets and the results are compared with traditional machine learning algorithms as well as different methods based on Naive Bayes, Bag-of-Words and Word2Vec. According to the experimental results, the proposed text classification model is 7-12% better than traditional machine learning algorithms.

In order to solve the problems of traditional traffic accident scene investigation, such as taking a long time, evidence easily lost and difficult to save in case of bad weather, low survey accuracy, and field measurement data, DJI Mavic 3E UAV is used to convert the collected data into digital two-dimensional ortho image and three-dimensional model by using DJI Intelligent map software, such as mid-way point flight, map construction aerial photography and oblique shooting. One-stop help traffic accident investigation comprehensively improve the efficiency of scene investigation, standard forensics, improve the accuracy of accident scene investigation, in order to quickly restore traffic order, ease the demand for police, and improve the identifiability, safety and timeliness of traffic accident scene investigation.

By improving the standard U-Net architecture, this paper proposes a novel semantic segmentation model, which incorporates multiple attention mechanisms to enhance the model’s capacity to capture multi-scale features. Specifically, we introduce the Efficient Multi-Scale Attention Module with CrossSpatial Learning (EMA), Spatial and Channel Squeeze and Excitation (SCSE), and Squeeze-andExcitation (SE) mechanisms into the standard U-Net network. These modules assist the network in learning significant information from feature maps at multiple scales while suppressing interference from irrelevant background. Experimental results demonstrate that incorporating attention mechanisms effectively enhances the prediction accuracy of the standard U-Net network for lane line semantic segmentation. The new model outperforms the standard U-Net model on our custom dataset, with particularly significant improvements in lane detection accuracy in scenarios with certain interference.