2021 11th International Conference on Information Science and Technology (ICIST)最新文献

{"title":"Detection of Small Size Defects in Belt Layer of Radial Tire Based on Improved Faster R-CNN","authors":"Pengcheng Li, Zihao Dong, Jianjie Shi, Zengzhi Pang, Jinping Li","doi":"10.1109/ICIST52614.2021.9440580","DOIUrl":"https://doi.org/10.1109/ICIST52614.2021.9440580","url":null,"abstract":"The cord structure of belt layer in radial tire is complex, and various defects such as cord overlapping, cord cracking and impurities may occur during manufacturing. With the development of deep learning, object detection based on convolutional neural network become a common defect detection method. Due to the variety of defect sizes, the receptive field of the feature map output by this kind of method is large, whereas the small-size defect yields weak feature and may be readily missed. In order to solve the problem that the feature extracted by the convolutional neural network of small-size defect is weak, a belt layer defect detection method based on improved Faster R-CNN is proposed. Faster-R-CNN’s convolution layers are used for feature fusion to solve the problem of insufficient feature extraction for subtle weeny defects. At the same time, Distance Intersection over Union (DIoU) is used to obtain object box that’s more sensitive to object scale to solve the problem of loose defect bounding boxes. The algorithm steps are as follows: Firstly, the belt layer area is segmented. We first segment the shoulder and belt layer areas by vertical projection, and then combine extreme value filtering (EVF) with binarization to segment the belt layer area according to the characteristics of the horizontal cord in the shoulder area. Secondly, construct the defect dataset of the belt layer and enlarge the area proportion of the defect target in the image. Thirdly, the shared convolutional layer of Faster R-CNN is used for front-layer feature fusion to ensure the feature map include higher-level features and higher resolution features. Finally, DIoU is used to get a bounding box that is more scale-sensitive. Experiments were conducted on the defect dataset containing 6316 object boxes for training and 1036 object boxes for test. Compared with the vanilla Faster R-CNN, the false negative rate decreased by 7.79%, the false positive rate decreased by 3.4%, the f-score improved by 5% and the detection box is more fitting for the defect object.","PeriodicalId":371599,"journal":{"name":"2021 11th International Conference on Information Science and Technology (ICIST)","volume":"54 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-05-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130980586","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Disturbance observer based on NTSM output tracking control for second-order systems with power integrators and input quantization","authors":"Yunxi Zhang, Haibin Sun","doi":"10.1109/ICIST52614.2021.9440571","DOIUrl":"https://doi.org/10.1109/ICIST52614.2021.9440571","url":null,"abstract":"In this paper, a composite finite-time tracking controller is designed for second-order systems with power integrators and input quantization, where the tracking reference signal is generated by a power-integrator exogenous system. A fixed-time disturbance observer is presented to estimate the unknown disturbances. And then a novel non-singular terminal sliding mode (NTSM) surface is constructed. Based on sliding mode surface (SMS) and disturbance estimation, two composite finite-time tracking controllers are developed for two different cases. The controllers make the system trajectories track the desired signal in finite-time. Simulation results are employed to demonstrate the effectiveness of the proposed schemes.","PeriodicalId":371599,"journal":{"name":"2021 11th International Conference on Information Science and Technology (ICIST)","volume":"94 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-05-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114789894","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Spatiotemporal Multi-Graph Convolutional Network for Taxi Demand Prediction","authors":"Genxuan Hong, Zhanquan Wang, Taoli Han, Hengming Ji","doi":"10.1109/ICIST52614.2021.9440573","DOIUrl":"https://doi.org/10.1109/ICIST52614.2021.9440573","url":null,"abstract":"Taxi demand prediction plays an important role in ITS (Intelligent Transportation System). This task is challenging due to the complex spatiotemporal correlations and semantic trends between different locations. Existing work tried to solve this problem by exploiting a variety of spatiotemporal models based on deep learning. However, we observe that more semantic pair-wise correlations among possibly distant roads are also critical for taxi demand prediction. To combine the spatiotemporal correlations with semantic correlations in the traffic network, this paper proposed an end-to-end framework called DeepTDP. First, we defined five kinds of spatial and semantic correlations, which are modeled into multi location graphs and fused by multi-graph convolutional network. Second, LSTM in encoder-decoder network is utilized to capture temporal correlation between future taxi demand values. Besides, a cross-entropy loss function based on error correction is designed to generate taxi demand predictions. Third, we apply a word embedding technique to reduce the dimension of decoded vector in output layer. Finally, we evaluate DeepTDP on two real world traffic datasets, the experiment results demonstrate effectiveness of our approach in comparison with variants of self and other baselines.","PeriodicalId":371599,"journal":{"name":"2021 11th International Conference on Information Science and Technology (ICIST)","volume":"086 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-05-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129028322","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A Deep Learning Based Method for Structuring the Chinese Pathological Reports of Lung Specimen","authors":"Tianzhong Lan, Jingwei Li, Xiuyuan Xu, Chengdi Wang, Zhang Yi, Wei-min Li, Jixiang Guo","doi":"10.1109/ICIST52614.2021.9440626","DOIUrl":"https://doi.org/10.1109/ICIST52614.2021.9440626","url":null,"abstract":"As a kind of electronic reports in text form, the Chinese pathology report of lung specimen contains a large amount of information that is important for clinicians to further analysis and mining. However, various expressions and no fixed format increases the difficulty of extracting and standardizing this information. In this paper, we focus on the extraction of lung lesion locations and the corresponding diagnosis from these reports. And to overcome the difficulties, a structured processing method based on deep learning and the idea of part-of-speech (POS) tagging was proposed. Firstly, the data of lung pathology specimen reports are preprocessed to normalize the medical terms. Secondly, the bidirectional Long Short-Term Memory (Bi-LSTM) neural network is adopted to extract the information of lesion locations and pathological diagnosis from each report. Finally, the obtained information is screened by an information filter method to generate the final structured results. Experimental results on the self-constructed datasets indicated that the proposed method can be beneficial for structuring pathology reports of lung specimen and obtained state-of-the-art results.","PeriodicalId":371599,"journal":{"name":"2021 11th International Conference on Information Science and Technology (ICIST)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-05-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129952151","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"FPGA Implementation of Object Detection Accelerator Based on Vitis-AI","authors":"Jin Wang, Shenshen Gu","doi":"10.1109/ICIST52614.2021.9440554","DOIUrl":"https://doi.org/10.1109/ICIST52614.2021.9440554","url":null,"abstract":"The emergence of YOLOv3 makes it possible to detect small targets. Due to the characteristics of the YOLO network itself, the YOLOv3 network has exceptionally high requirements for computing power and memory bandwidth and it usually needs to be deployed on a dedicated hardware acceleration platform. FPGAs is a logically reconfigurable hardware chip with substantial advantages in terms of performance and power consumption, so it is a good choice to deploy a deep convolutional network. In the research of this paper, we proposed a reconfigurable YOLOv3 FPGA hardware accelerator based on the AXI bus ARM+FPGA architecture. The YOLOv3 network quantifies through Vitis AI, and a series of operations such as model compression and data pre-processing can save accelerator chips and the access time of external storage. Pipeline operation enables FPGAs to achieve higher throughput. Compared with the GPU implementation of the YOLOv3 model, it is found that the hardware implementation of the FPGA-based YOLOv3 accelerator has lower energy consumption and can achieve higher throughput.","PeriodicalId":371599,"journal":{"name":"2021 11th International Conference on Information Science and Technology (ICIST)","volume":"19 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-05-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123485969","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Fully Circuit Implementation of a two-layer Memristive Neural Network for Pattern Recognition","authors":"Mian Li, Xiaoping Wang, Zhanfei Chen","doi":"10.1109/ICIST52614.2021.9440557","DOIUrl":"https://doi.org/10.1109/ICIST52614.2021.9440557","url":null,"abstract":"In this paper, a fully circuit implementation of Memristive Neural Network (MNN) is proposed. The forward calculation of the network is based on winner-take-all (WTA) mechanism. The weight updating is achieved through the difference of pre-spike and post-spike, which is more close to the biological weight adjustment mechanism. The network is implemented in a full circuit without additional control units. The designed circuit consists of four modules. The memristive crossbar array module with one-memristor (1M) unit structure can effectively calculate the vector-matrix multiplication with only one step. The switch S is replaced by the transistor in the designed leaky-integrate-and-fire (LIF) module, which can control the integration and leakage of the membrane voltage and realize the lateral inhibition between output neurons. Connecting the integrated monostable trigger and the difference circuit, the post-spike generating module can output the required post-spike. The signal switch module realizes the switching of signals connected to memristors by using voltage-controlled switches. The combination of two modules validly realizes weight updating. The functions of four modules are verified separately. We performed a simulation experiment of 5×3 pixels image classification based on the designed circuit in PSPICE. The circuit output results and the high classification accuracy prove the circuit can be effectively applied in pattern recognition. The noise experiment shows the robustness of the designed circuit.","PeriodicalId":371599,"journal":{"name":"2021 11th International Conference on Information Science and Technology (ICIST)","volume":"39 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-05-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128358924","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Fast Generation of Optimal Topology for 3D Wireless Sensor Networks","authors":"Guoqiang Wang, Yuanyuan Jiang, Xin Cao, Xiaoduo Li, He Luo","doi":"10.1109/ICIST52614.2021.9440621","DOIUrl":"https://doi.org/10.1109/ICIST52614.2021.9440621","url":null,"abstract":"The topology control for three-dimensional (3D) wireless sensor network is a basic problem in the research of 3D wireless sensor network, and it is also one of the important supporting technologies. Aiming at the topology control problem of three-dimensional wireless sensor networks, this paper proposes an optimal topology rapid generation algorithm based on minimum spanning tree and edge addition operation, and analyzes the time complexity of the algorithm. Compared with the existing algorithms, this algorithm can generate a three-dimensional optimal rigid graph as the wireless sensor network topology in a shorter time to ensure the connectivity and robustness of the network. Finally, the effectiveness of the algorithm is further verified by comparing simulation experiments.","PeriodicalId":371599,"journal":{"name":"2021 11th International Conference on Information Science and Technology (ICIST)","volume":"162 3 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-05-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129150806","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Research on Simplified Mesh Deformation Based on Differentiable Computation","authors":"Zhuo Shi, Shuzhen Zeng, Xiaonan Luo","doi":"10.1109/ICIST52614.2021.9440622","DOIUrl":"https://doi.org/10.1109/ICIST52614.2021.9440622","url":null,"abstract":"In this paper, we propose a simplified mesh deformation method based on the differentiable calculation and uses the Pytorch3D library of deep learning. There are four stages, simplification, deformation, subdivision, re-deformation in this method. The simplification stage transforms the original target mesh into a simple mesh. The deformation stage uses the Pytorch3D tool to predict the simple mesh in the simplification result. The subdivision stage subdivides the resulting mesh of deformation, and the re-deformation stage uses the subdivision stage result mesh as the source mesh to predict the original target mesh. Our experiment shows that the number of iterations is similar or less in terms of shape and local features after simplifying the predicted target mesh. Our method is superior to the direct mesh deformation method in terms of mesh deformation speed and local mesh characteristics of deformation and has a better deformation effect.","PeriodicalId":371599,"journal":{"name":"2021 11th International Conference on Information Science and Technology (ICIST)","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-05-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126658793","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Alzheimer’s disease diagnosis method based on convolutional neural network using key slices voting","authors":"Zhongyi Hu, Qi Wu, Changzu Chen, Lei Xiao, Sha Jin","doi":"10.1109/ICIST52614.2021.9440595","DOIUrl":"https://doi.org/10.1109/ICIST52614.2021.9440595","url":null,"abstract":"With the wide application of convolutional neural networks in the field of computer vision, its application in medical image analysis has become a research hotspot. Due to the significant social impact of Alzheimer’s disease, the detection of Alzheimer’s disease in magnetic resonance imaging data has become the focus of research. At present, researchers have found that a large number of studies have data leakage problems, resulting in a poor generalization of the training model. On the Alzheimer’s Disease Neuroimaging Initiative database, this paper conducts comparative experiments based on the random split of slices and the independent split of subjects and verifies the existence of data leakage problem. Furthermore, based on the independent data slices of the subjects, this paper innovatively proposes an auxiliary diagnosis method of Alzheimer’s disease based on the slice voting algorithm and uses different slice intervals of the coronal plane to train the model, as well as 3D magnetic resonance imaging data to train the 3D model convolutional neural network model, experimental results show that the accuracy of the proposed method is 93.10%, which is 5.39% higher than the slice recognition rate, and 4.59% higher than that of 3D magnetic resonance imaging. At the same time, the experimental results show that the slice interval of [75, 106] has the best effect on the diagnosis of Alzheimer’s disease. The experimental results of this paper have an important role in the auxiliary diagnosis of Alzheimer’s disease.","PeriodicalId":371599,"journal":{"name":"2021 11th International Conference on Information Science and Technology (ICIST)","volume":"100 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-05-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114947289","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}