Computer and Modernization

Multiple Unmanned Aerial Vehicles Three-dimensional Cooperative Route Planning Based on Improved GWO Algorithm

JIAO Jian, JI Yuanfa, SUN Xiyan, WU Jianhui, LIANG Weibin

2024, 0(10): 1-6. doi:10.3969/j.issn.1006-2475.2024.10.001

Asbtract ( 103 )

PDF (2694KB) ( 106 )

References | Related Articles | Metrics

To overcome the problems of poor cooperation， immersing local minimization， low convergence speed and poor solving accuracy in solving the collaborative route by GWO algorithm for multiple unmanned aerial vehicles， an improved GWO-based three-dimensional collaborative route planning algorithm for multiple unmanned aerial vehicles is proposed. Firstly， a three-dimensional collaborative trajectory planning mathematical model for multiple unmanned aerial vehicles is established， using the weighted sum of consumption cost， height cost， threat cost， spatial constraint， time constraint， and penalty term as the objective function. Secondly， the Greedy algorithm and Tent mapping are combined to improve the fitness of the population and preserve the diversity of the population to reduce the possibility of falling into local optima； then we optimize the convergence factor to improve the rate of convergence of the algorithm. Afterwards， we design a dynamic weight position update method to enhance the exploration and development capabilities of the algorithm. Finally， the improved GWO algorithm is applied to solve the trajectory planning problem of multiple unmanned aerial vehicles， and compared with GWO algorithm and CSGWO algorithm. The simulation results indicate that the proposed improved GWO algorithm enhance the solution accuracy by 64.8% and 16.7%， as well as the convergence speed by 28.5% and 25.4%， respectively. Additionally the synergy ability is significantly better than that of the comparison algorithms.

Semantic Segmentation Algorithm for Rainy Road Scene and Its Mobile Deployment

ZHOU Anda, TANG Chaoying

2024, 0(10): 7-13. doi:10.3969/j.issn.1006-2475.2024.10.002

Asbtract ( 72 )

PDF (2649KB) ( 80 )

References | Related Articles | Metrics

The existing semantic segmentation models are susceptible to the interference of raindrop occlusion， and the performance is poor on the rainy road scene dataset. Moreover， they did not focus on the two important categories of vehicles and pedestrians in the road scene. Aiming at the above two problems， this paper designs a semantic segmentation algorithm for rainy road scenes and deploys it on a mobile terminal to promote the development of autonomous driving technology. A fast fusion pyramid pooling module is proposed to make the feature map integrate rich global semantic information and local detail information， and effectively segment the raindrop obscured scene. A multiple attention fusion module is proposed， and the category weight of vehicles and pedestrians in the loss function is increased to enhance the model’s attention to vehicles and pedestrians. The model is deployed to the mobile terminal with the help of Android Studio platform， and the ONNX Runtime is used for forward inference， and the segmentation effect is consistent with that of the computer terminal. Compared with five recent models on the Rainy WCity dataset， the segmentation accuracy of this model is the same on the computer terminal and the mobile terminal. Specifically， PA and mIoU are 95.25% and 72.96%， vehicle PA and IoU are 84.04% and 74.15%， and pedestrian PA and IoU are 34.91% and 26.37%， respectively， which are higher than those of the other five models. In addition， the FPS of the model in the computer and mobile terminals are 45.46 and 1.26， respectively， and the segmentation speed is fast. The model proposed in this paper can effectively segment the road scene image under the shelter of rain on the mobile terminal， and it is more accurate to segment vehicles and pedestrians.

Grounding Grid Corrosion Localization Based on Improved Sparrow Search Algorithm

YANG Zhengke, SHEN Xiaodong, WANG Kaixiang, HE Li

2024, 0(10): 14-20. doi:10.3969/j.issn.1006-2475.2024.10.003

Asbtract ( 69 )

PDF (1982KB) ( 73 )

References | Related Articles | Metrics

The grounding grid is an important component of the normal operation of the power system， the resistance value of the grounding grid is related to system stability， safety protection， and current setting directly. The materials of the grounding grid are mainly ordinary carbon steel or galvanized carbon steel and perennially underground which is prone to corrosion and failure. When the corrosion and failure of grounding grid happens， it is necessary to locate and repair corrosion faults timely. With the rapid development of electrical systems， the performance requirements of grounding grids have become stricter， which puts forward higher requirements for the precision and accuracy of grounding grid corrosion localization. In response to the current situation of low accuracy in locating corrosion of grounding grids， this article proposes an improved sparrow search algorithm for corrosion localization of grounding grids (INLSSA). The algorithm uses ICMIC chaotic mapping， integrates the northern goshawk optimization exploration phase position strategy， and incorporates Levy flight disturbances in the following phase to improve the sparrow search algorithm. We establish a fault diagnosis model based on electrical network theory and micro processing method. Finally， we use INLSSA algorithm for solution. The experimental results show that compared with sparrow search algorithm， northern goshawrk optimization algorithm， grey wolf optimizer， dung beetle optimizer， golden jackal optimization， INLSSA can effectively locate faults and has high stability and accuracy， which can be used as a reference for actual grounding grid corrosion location.

Visual SLAM Loop Closure Detection Algorithm Based on Improved MobileNetV3

YANG Jun1, HU Wei1, ZHU Wenfu2

2024, 0(10): 21-26. doi:10.3969/j.issn.1006-2475.2024.10.004

Asbtract ( 65 )

PDF (2129KB) ( 68 )

References | Related Articles | Metrics

To address the inaccuracies in loop closure detection by traditional algorithms under variable lighting， with dynamic objects， and changing viewpoints， leading to the problem of large error in robot mapping， this paper introduces an algorithm using an enhanced MobileNetV3 for visual SLAM. The work improves the Coordinate Attention mechanism within the feature extraction network， enhancing spatial information extraction to meet loop detection needs. Features are then dimensionally reduced via an autoencoder and assessed for similarity to detect loop closures. Experimental results on the City Centre dataset indicate a 21.8 percentage points increase in detection accuracy and a significant speed improvement compared with traditional methods. This approach also more effectively reduces cumulative errors in visual SLAM systems， ensuring greater real-time performance.

Improved Roadside Monocular View Small Target Detection Algorithm Based on YOLOv5

WEI Xuecheng1, JIANG Lingyun1, LI Yan2, HE Fei2

2024, 0(10): 27-34. doi:10.3969/j.issn.1006-2475.2024.10.005

Asbtract ( 78 )

PDF (2402KB) ( 105 )

References | Related Articles | Metrics

Aiming at the problems of low recognition accuracy and fewer features of long-distance targets and small targets in the roadside view under vehicle-road cooperative sensing traffic scenarios， an improved algorithm for small target detection based on YOLOv5 is proposed. Firstly， in the backbone network， the GAM attention module is added to enhance the feature extraction ability of the network. Secondly， RepBi-PAN is introduced to replace the PANet structure of the original neck network to increase the network’s ability to localize small targets. Finally， the use of SIoU loss function instead of the original CIoU loss function can effectively avoid the arbitrary matching of the prediction frames in the regression process， thus enhancing the robustness of the model and accelerating the training speed of the network model. The experimental results show that compared with the original YOLOv5 6.0 version， the average accuracy mAP of each category is improved by 6.9 percentage points when the intersection over union IoU is 0.5， and the average accuracy mAP of each category is improved by 6.4 percentage points when the intersection over union IoU is 0.95， which effectively improves the detection capability of small target detection in the road-side view.

Enhanced Detection and Evaluation of Facial Wrinkles Based on Image Denoising

ZHONG Jiaxuan1, LANG Xun1, ZHANG Ningtao2, ZHANG Zhao3, GUO Zhenyu2, ZHANG Mei2, ZHANG Yufeng1

2024, 0(10): 35-41. doi:10.3969/j.issn.1006-2475.2024.10.006

Asbtract ( 70 )

PDF (3371KB) ( 49 )

References | Related Articles | Metrics

Facial skin wrinkles are positively correlated with physiological age and are an important feature of aging. Existing wrinkle detection algorithms are influenced by facial features and image backgrounds， and can only detect specific regions of the face. Moreover， they focus heavily on the detection of horizontal forehead wrinkles， and suffer from inaccurate localization and identification of vertical or horizontal discontinuous textures as wrinkles， leading to low detection accuracy. In addition， in terms of wrinkle evaluation， existing methods or indicators cannot achieve quantitative evaluation of the overall wrinkles of the human face. To solve above problems， an enhanced facial wrinkle detection and evaluation method based on image denoising is proposed. Firstly， the facial image is preprocessed （denoised） using 2D-VMD to reduce the undesirable effects of non-wrinkled regions. Then， the hybrid Hessian filter is used to locate the wrinkle regions. Furthermore， the Dlib library is employed to eliminate facial features and image backgrounds， enabling wrinkle detection for the entire face. Finally， an improved quantitative evaluation method for wrinkles is proposed based on the geometric and intensity constraints of the wrinkle curve object. This method is not limited to the evaluation of a single wrinkle， which fills the gap in the quantitative evaluation of the overall facial wrinkles. The effectiveness of the proposed method is verified on representative 2D facial images of human faces.

Pipelines in Drawings Detection Method Based on Improved Mask R-CNN and LSD

HUANG Shanshan1, WU Wei2, XU Yuqing1, WEI Jie1

2024, 0(10): 42-48. doi:10.3969/j.issn.1006-2475.2024.10.007

Asbtract ( 46 )

PDF (4156KB) ( 54 )

References | Related Articles | Metrics

Aiming at the problems of poor precision of pipelines detection， false detection and missed detection caused by indistinct pipelines features， large differences in pipeline scales and pipeling intersections in the nuclear power axonometric drawings，an method for pipelines detection based on improved Mask R-CNN and LSD is proposed. Firstly， aiming at the problems of indistinct pipelines features， the recognition target is adjusted from the pipelines to the pipelines and its dimensioned lines. The recognition target geometry features are added. Secondly， the Mask R-CNN network is improved， and the BiFPN structure is used to enhance the ability to extract target features at different scales. We change the original NMS to DIoU-NMS to improve the accuracy of intersecting pipelines detection. Finally， the LSD algorithm is used to detect the lines in the target image， and then the pipeline lines are obtained by conditional constraint filtering and least square fitting. The experimental results show that the improved Mask R-CNN algorithm can well solve the problems of missed detection and false detection， and its accuracy recognition rate reaches 90.04%. Combining LSD line detection， conditional constraint， and least squares fitting algorithm， pipeline lines are obtained， which meets the requirements of pipelines detection in the drawings.

Construction Protective Wear Detection Based on Improved YOLOv5

HAN Ruichao, MENG Lingjun, AO Licheng, XIE Yubin, ZHEN Mingshuo

2024, 0(10): 49-54. doi:10.3969/j.issn.1006-2475.2024.10.008

Asbtract ( 78 )

PDF (2926KB) ( 62 )

References | Related Articles | Metrics

A protective equipment detection method based on CAS-YOLOv5 is proposed to address the issues of missed detection， positioning errors， and low accuracy in the detection of helmets and safety vests in complex environments using existing algorithms. Firstly， in order to solve the problem of missed detection of small targets， the ASFF（Adaptively Spatial Feature Fusion） detection head is used to improve the model’s recognition ability for small targets. Secondly， in order to improve the detection accuracy of the model and correct positioning errors， a coordinate attention mechanism is added to the backbone network to enhance the model’s perception of important target areas and improve the recall rate of target detection. Once again， we use the WIoU loss function to accelerate the convergence speed of model training， and add Slim-Neck composed of GSConv（Group Shuffle Convolution） modules at the network neck to reduce the dimensionality of feature maps and improve the computational efficiency of the model. Finally， through ablation and comparative experiments on a public dataset， the mAP index of this method is improved by 5.6 percentage points， and the recall rate is increased by 4 percentage points compared to the YOLOv5 model. The improved method can reduce the missed detection rate and effectively improve detection performance， which has good application prospects in construetion protective equipent detection.

Domain Adaption-based Underwater Image Enhancement Algorithm

DU Feiyu, WANG Haiyan, YAO Haiyang, CHEN Xiao

2024, 0(10): 55-60. doi:10.3969/j.issn.1006-2475.2024.10.009

Asbtract ( 58 )

PDF (2333KB) ( 63 )

References | Related Articles | Metrics

Underwater image enhancement is a key technology for underwater missions. Aiming at the problems of color distortion and image blurring in underwater images， this paper designs an underwater domain adaptation network (UDA Net) based on the domain adaptive method to achieve effective enhancement of underwater images under unsupervised conditions. The sharpness of the original underwater images is significantly improved. Based on U-Net network framework， the algorithm uses convolutional neural network and multi-head attention mechanism for feature extraction， introduces adversarial learning idea， and adds discrimination network to domain feature extraction module and output module. Meanwhile， it optimizes feature enhancement loss， feature alignment loss and output alignment loss in the source domain to ensure style transfer and feature alignment from the source domain to the target domain， achieving underwater enhancement. In addition， the public underwater data sets EUVP， UIEB and UFO-120 are used for experimental verification， and the experimental results are compared with cutting-edge enhancement algorithms. The effectiveness of UDA Net algorithm is proved， and it has a good application prospect in underwater image enhancement tasks.

Power Information Data Fusion Model Based on Improved Extreme Learning Algorithm

DU Mengjun1, LI Ang1, TONG Jun1, QIAN Jin1, KANG Kai1, WANG Ruoding1, JIN Wenxing2

2024, 0(10): 61-64. doi:10.3969/j.issn.1006-2475.2024.10.010

Asbtract ( 44 )

PDF (2671KB) ( 65 )

References | Related Articles | Metrics

In view of the poor interaction ability of power flow and information flow in power communication network， which leads to the low utility of power information data and can not meet the needs of the actual new power system， this study proposes an improved extreme learning algorithm based on communication data fusion method to improve the performance of power information data acquisition and energy-efficient data processing in the new-type power system. Firstly， the low rank autoregressive tensor completion （LATC） algorithm is used to integrate the multi-source heterogeneous data transmitted by power flow and information flow back to the terminal， and reduce the impact of missing data. Further， the extreme learning machine （ELM） algorithm is used to construct the relationship between the data as data features， and the feature set is output to complete the data feature level fusion. Then， in order to improve the fusion accuracy in the fusion task， attention mechanism is added to the extreme learning machine as the underlying infrastructure. Finally， the experimental results show the effectiveness of the method.

Environmental Topology Task Scheduling Based on Diverse Hierarchical Difference Optimization Genetic Algorithm

WANG Jia1, GU Wenjun1, JU Weigang2, LI Yuwei1, ZHANG Yunlong2, MI Chuanmin3, ZHOU Zhipeng3

2024, 0(10): 65-73. doi:10.3969/j.issn.1006-2475.2024.10.011

Asbtract ( 53 )

PDF (1677KB) ( 43 )

References | Related Articles | Metrics

Under the background of the deep promotion of the “East-West Computing Requirement Transfer” project in China， the deployment and scheduling of the environment in the computing power network center faces many challenges， such as the uncertainty of the number， size， topology complexity， dependency constraints， and network transmission volume of the environment. This paper proposes a diverses hierarchical difference optimization genetic algorithm （DHDO-GA） to solve these problems. DHDO-GA aims at optimizing the task execution span makespan and resource utilization rate， while considering the load balancing of resources. In order to guide the entire population to quickly converge to the global optimal solution， DHDO-GA distributes chromosomes at different hierarchical levels based on fitness value and similarity， and abstracts and clusters them into elite populations. Simulation experiments show that the DHDO-GA algorithm is superior to traditional genetic algorithms and several improved genetic algorithms， with greater advantages in terms of search capability， algorithm stability， and result quality and reliability.

Fine-grained Image Classification Based on Res2Net and Recursive Gated Convolution

WANG Yingying, HAO Xiao

2024, 0(10): 74-79. doi:10.3969/j.issn.1006-2475.2024.10.012

Asbtract ( 70 )

PDF (1792KB) ( 60 )

References | Related Articles | Metrics

Extracting discriminative regions in images plays a crucial role in fine-grained image classification. Existing fine-grained image classification methods ignore the multi-scale information of the image and the interaction of adjacent spatial position information，and it is difficult to accurately extract subtle features. Moreover， the traditional CNN method is insufficient to capture long-distance semantic information and cannot obtain accurate global information.To address these issues， a fine-grained classification algorithm based on Res2Net and recursive gated convolution module is designed. In this network， the weakly supervised data augmentation network （WS-DAN） is used for data expansion to prevent overfitting， and Res2Net is used as a feature extraction network， which can extract image information of different scales， increase the receptive field of network layer. Meanwhile， a recursive gated convolution module is introduced into the network to further fuse information and realize high-order feature interaction to improve network modeling capabilities. The proposed method achieves 90.36%， 93.1% and 94.3% accuracy on the three public datasets of CUB-200-2011， Stanford Dogs and FGVC-Aircraft， respectively， which can effectively extract subtle features of images and achieve classification.

National Path Planning Algorithm Based on Embedded System

ZHANG Yongliang1, 2, WANG Jiarun2

2024, 0(10): 80-86. doi:10.3969/j.issn.1006-2475.2024.10.013

Asbtract ( 50 )

PDF (4753KB) ( 42 )

References | Related Articles | Metrics

The memory resources provided by embedded systems are very limited. In the calculation of large memory， the problem of low calculation efficiency or even crash often occurs. For this problem， a practical algorithm for nationwide path planning based on embedded systems is designed. The algorithm uses the domestic four-dimensional navigation data as the basic data for generating the navigation planning data model， including: the topological relationship of the basic computing nodes in the sub-map， the sequence list of traffic taboo information， and the connection information between nodes and roads. According to the characteristics of the embedded system， we propose the design based on the idea of key node enrichment to extract the national basic navigation road network， the data scheduling strategy for real-time planning and calculation， and the optimization of the practical two-way Dijkstra algorithm based on the embedded system. The experimental results show that the algorithm in this paper has good performance on embedded systems with limited computing power and computing memory.

Enhanced Indoor Positioning Method for VSLAM Based on Object Recognition

SHEN Junjie, NIE Yun, WANG Guowei

2024, 0(10): 87-92. doi:10.3969/j.issn.1006-2475.2024.10.014

Asbtract ( 66 )

PDF (2573KB) ( 57 )

References | Related Articles | Metrics

In response to the low spatial positioning accuracy and insufficient robustness in indoor dynamic scenarios， this paper proposes an enhanced positioning method suitable for indoor dynamic environments. Firstly， common indoor objects are categorized based on their motion properties， and object detection is performed using the YOLOv5s neural network to obtain the positions of the target detection boxes for subsequent dynamic feature point screening. Then， a feature point selection strategy is designed， which uses edge detection and depth information filtering to determine which feature points within the target detection boxes have the potential for dynamic motion. Finally， a keyframe selection algorithm that integrates time step and the number of feature points is proposed to eliminate redundant keyframes and reduce feature information overlap between multiple frames. The proposed positioning enhancement method is transplanted into ORB-SLAM2 and tested based on the publicly available RGB-D dataset from the Technical University of Munich （TUM）. The experimental results show that the average positioning error has reduced compared to ORB-SLAM2， validating the effectiveness of the proposed method.

Object Detection Models Distillation Technique for Industrial Deployment

SHI Xingyu1, LI Qiang2, ZHUANG Li3, LIANG Yi3, WANG Qiulin3, CHEN Kai3, WU Chenzhou3, CHANG Sheng1

2024, 0(10): 93-99. doi:10.3969/j.issn.1006-2475.2024.10.015

Asbtract ( 59 )

PDF (1549KB) ( 48 )

References | Related Articles | Metrics

The application scenarios of deep learning object detection models are quite extensive. However， the detection accuracy of deployed models is often low due to the performance limitations of deployment devices. To enhance the performance of detection models， this paper proposes an efficient dynamic distillation training method. This method innovatively introduces a dynamic sample assignment strategy to select high-quality outputs of the teacher model， and pairs this with dynamic weight adjustment of distillation loss， thereby improving the traditional distillation algorithm used in object detection models. Experimental results on a dataset for electrical grid safety construction indicate that， compared to direct training， this method increased the Average Precision （AP） value of the YOLOv6-n model by an average of 2.63 percentage points. The distillation method proposed in this paper does not affect the inference speed of the original deployment model and helps to enhance the detection performance of object detection models in various industrial scenarios.

A Review of Bionic Olfactory Model Construction Methods and Applications

LI Wei1, HE Haobo1, LI Guang2, KUANG Lidan1, ZHANG Jin1

2024, 0(10): 100-106. doi:10.3969/j.issn.1006-2475.2024.10.016

Asbtract ( 74 )

PDF (1163KB) ( 55 )

References | Related Articles | Metrics

Olfaction plays a crucial role in the survival and reproduction of organisms. Bio-inspired olfactory models， as a significant approach to studying smell， have the potential to drive the development of various fields such as biology， neuroscience， environmental monitoring， computer science， and medicine. In addressing the challenges presented by the numerous existing bio-inspired olfactory models， this work initially conducts a comprehensive review of the composition and operation principles of the biological olfactory nervous system， providing a theoretical foundation for constructing bio-inspired olfactory models. Secondly， based on the principle of simulating the binding process between olfactory receptor cells and odor molecules， the methods for constructing bio-inspired olfactory models are meticulously categorized and summarized. Each method is thoroughly dissected in terms of its working mechanisms and principles， and their strengths and limitations in bio-inspired olfactory model research are summarized. Finally， addressing the issues encountered during the construction of bio-inspired olfactory models， feasible recommendations are provided based on theoretical derivations and practical application experiences.

Label Recommendation Methods for Public Cultural Resources

JIAO Yikai1, 2, ZHU Xinjuan1, 2

2024, 0(10): 107-112. doi:10.3969/j.issn.1006-2475.2024.10.017

Asbtract ( 47 )

PDF (2181KB) ( 59 )

References | Related Articles | Metrics

Resource labels play an indispensable role in the era of information explosion， and the use of labels can greatly reduce the workload of recommendation systems and improve their accuracy. A public cultural resource recommendation method based on the relevance of resource labels is designed based on the resources of the national public cultural cloud platform. Firstly， the Integrating Global Semantics BERT TextCNN Model is proposed， which extracts the deep semantic relationships between local and global resource texts and labels to obtain the text correlation between resources and labels. Secondly， the keyword correlation between resources and labels is mined based on the TF-IDF algorithm. Finally， the correlation between resources and labels is obtained by using perceptron model and the recommended sequence of public cultural label resources is ultimately obtained. Multigroup experiments are conducted on the Reuters-21578 and National Public Culture Cloud datasets. The experiments results show that the resource recommendation effect of our method is superior to the baseline model.

Sentiment Consistency Detection Based on Cross Modal Attention Fusion and#br# Information Perception

YANG Shijun1, DI Guangyi1, GAO Jun1, CHEN Jianfei1, WANG Yaokun1, JI Xiaohan2

2024, 0(10): 113-119. doi:10.3969/j.issn.1006-2475.2024.10.018

Asbtract ( 45 )

PDF (1387KB) ( 51 )

References | Related Articles | Metrics

With the rapid development of information technology， massive amounts of image and text information are constantly generated and disseminated through various channels. The recognition and detection technology for multimodal data is widely used in many fields such as e-commerce， healthcare， logistics， finance， and construction. Sentiment consistency detection aims to explore how to accurately determine whether sentiments expressed in different modal data are consistent. Most existing sentiment consistency detection models usually adopt implicit fusion， without explicitly aligning sentiments between modalities， and ignoring the important role of sentiment words in detection. Therefore， a model is proposed based on cross-modality attention fusion and information perception for sentiment consistency detection. The model utilizes a dual channel module based on BERT to capture the dynamic interaction between image and text modalities， introduces external knowledge to enhance text representation， aggregates image and text based on sentiment information， builds a common attention matrix to capture the uncoordinated features between text sentences and text labels， as well as between the sentiment vectors of text sentences and text labels， and improves the accuracy of sentiment consistency detection between image and text. The experimental results on a public multi-modal dataset based on X（former Twitter）demonstrates the superiority of the proposed model.

An Aggregation Model of Historical Archive Resources Based on Multimodal#br# Information Fusion

WANG Cong1, YANG Wenjuan1, DING Xingwang2

2024, 0(10): 120-126. doi:10.3969/j.issn.1006-2475.2024.10.019

Asbtract ( 59 )

PDF (2062KB) ( 47 )

References | Related Articles | Metrics

Restricted by the diversity， individuation and value characteristics of historical archive resources， the aggregation results obtained by the traditional aggregation model have a large deviation from the expected effect， and there are serious resource losses in the process of some historical resources aggregation， and its aggregation accuracy is difficult to meet the requirements of the aggregation scene. In order to improve the application of traditional aggregation model in historical archive resources， a multi-mode information fusion algorithm is introduced to build a historical archive resource distribution model based on multi-mode information fusion. Firstly， the correlation distribution of resource features is sorted out， and the semantic aggregation of historical archive resources is set. Then， the standard of effective relation parameters in the polymerization process and the polymerization method are fixed. Finally， the model is obtained through aggregation of historical file resources based on multi-modal information fusion. Through the feasibility simulation data verification with the other two aggregation models， it is shown that each aggregation index of the proposed model can meet the aggregation conditions of historical archive resources， and the model has stable performance and high reliability， which is suitable for small-scale application at present， and has high value for promotion and in-depth research.

Table of Content